├── evaluation_scripts ├── __init__.py ├── converters.py └── evaluation_atis_snips.py ├── data ├── ATIS │ └── README.md ├── SNIPS │ └── README.md └── NLU-Evaluation-Data │ └── README.md ├── .gitignore ├── configs ├── table-2 │ ├── intent-classification │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ └── entity-recognition │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml ├── table-3 │ ├── ConveRT │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── ConveRT-mask-loss │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── GloVe │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── GloVe-mask-loss │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── BERT │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-10.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ └── config-NLU-Evaluation-Data-Fold-9.yml │ ├── BERT-mask-loss │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── sparse │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── sparse-mask-loss │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── sparse-ConveRT │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── sparse-ConveRT-mask-loss │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── sparse-GloVe │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── sparse-GloVe-mask-loss │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml │ ├── sparse-BERT │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-10.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ └── config-NLU-Evaluation-Data-Fold-9.yml │ └── sparse-BERT-mask-loss │ │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ │ └── config-NLU-Evaluation-Data-Fold-10.yml ├── table-5 │ ├── config-ATIS-sparse.yml │ ├── config-SNIPS-sparse.yml │ ├── config-ATIS-sparse-ConveRT.yml │ ├── config-SNIPS-sparse-ConveRT.yml │ ├── config-ATIS-sparse-GloVe.yml │ └── config-SNIPS-sparse-GloVe.yml ├── table-1 │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ └── config-NLU-Evaluation-Data-Fold-10.yml └── table-4 │ ├── config-NLU-Evaluation-Data-Fold-1.yml │ ├── config-NLU-Evaluation-Data-Fold-2.yml │ ├── config-NLU-Evaluation-Data-Fold-3.yml │ ├── config-NLU-Evaluation-Data-Fold-4.yml │ ├── config-NLU-Evaluation-Data-Fold-5.yml │ ├── config-NLU-Evaluation-Data-Fold-6.yml │ ├── config-NLU-Evaluation-Data-Fold-7.yml │ ├── config-NLU-Evaluation-Data-Fold-8.yml │ ├── config-NLU-Evaluation-Data-Fold-9.yml │ └── config-NLU-Evaluation-Data-Fold-10.yml ├── README.md └── run.sh /evaluation_scripts/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /data/ATIS/README.md: -------------------------------------------------------------------------------- 1 | Data taken from https://github.com/MiuLab/SlotGated-SLU -------------------------------------------------------------------------------- /data/SNIPS/README.md: -------------------------------------------------------------------------------- 1 | Data taken from https://github.com/MiuLab/SlotGated-SLU -------------------------------------------------------------------------------- /data/NLU-Evaluation-Data/README.md: -------------------------------------------------------------------------------- 1 | Data taken from https://github.com/xliuhw/NLU-Evaluation-Data 2 | -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- 1 | .idea 2 | 3 | # Byte-compiled / optimized / DLL files 4 | __pycache__/ 5 | *.py[cod] 6 | *$py.class 7 | 8 | # experiment results 9 | experiments/* 10 | 11 | .DS_Store 12 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | 16 | data: 17 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 18 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 19 | -------------------------------------------------------------------------------- /configs/table-2/intent-classification/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | entity_recognition: False 8 | intent_classification: True 9 | use_masked_language_model: False 10 | number_of_transformer_layers: 0 11 | hidden_layers: 12 | text: [256,128] 13 | weight_sparsity: 0 14 | 15 | data: 16 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 17 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 18 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: False 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "DIETClassifier" 7 | epochs: 200 8 | batch_size: [64, 128] 9 | entity_recognition: True 10 | BILOU_flag: Flase 11 | intent_classification: True 12 | use_masked_language_model: True 13 | transformer_size: 256 14 | number_of_transformer_layers: 2 15 | number_of_attention_heads: 4 16 | hidden_layers_sizes: 17 | text: [] 18 | label: [] 19 | dense_dim: 20 | text: 512 21 | label: 20 22 | scale_loss: False 23 | use_sparse_input_dropout: True 24 | weight_sparsity: 0 25 | 26 | data: 27 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 28 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 29 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: False 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "DIETClassifier" 8 | epochs: 200 9 | batch_size: [64, 128] 10 | entity_recognition: True 11 | BILOU_flag: Flase 12 | intent_classification: True 13 | use_masked_language_model: True 14 | transformer_size: 256 15 | number_of_transformer_layers: 2 16 | number_of_attention_heads: 4 17 | hidden_layers_sizes: 18 | text: [] 19 | label: [] 20 | dense_dim: 21 | text: 512 22 | label: 20 23 | scale_loss: False 24 | use_sparse_input_dropout: True 25 | weight_sparsity: 0 26 | 27 | data: 28 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 29 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 30 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: False 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-5/config-ATIS-sparse.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: True 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/ATIS/train.md" 35 | test_file: "data/ATIS/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-5/config-SNIPS-sparse.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: True 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/SNIPS/train.md" 35 | test_file: "data/SNIPS/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/BERT-mask-loss/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "DIETClassifier" 10 | epochs: 200 11 | batch_size: [64, 128] 12 | entity_recognition: True 13 | BILOU_flag: Flase 14 | intent_classification: True 15 | use_masked_language_model: True 16 | transformer_size: 256 17 | number_of_transformer_layers: 2 18 | number_of_attention_heads: 4 19 | hidden_layers_sizes: 20 | text: [] 21 | label: [] 22 | dense_dim: 23 | text: 512 24 | label: 20 25 | scale_loss: False 26 | use_sparse_input_dropout: True 27 | weight_sparsity: 0 28 | 29 | data: 30 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 31 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 32 | -------------------------------------------------------------------------------- /configs/table-5/config-ATIS-sparse-ConveRT.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: True 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/ATIS/train.md" 36 | test_file: "data/ATIS/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-5/config-SNIPS-sparse-ConveRT.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: True 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/SNIPS/train.md" 36 | test_file: "data/SNIPS/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-5/config-ATIS-sparse-GloVe.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: True 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/ATIS/train.md" 37 | test_file: "data/ATIS/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-5/config-SNIPS-sparse-GloVe.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: True 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/SNIPS/train.md" 37 | test_file: "data/SNIPS/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: False 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-3/sparse-mask-loss/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "WhitespaceTokenizer" 5 | - name: "CountVectorsFeaturizer" 6 | analyzer: "word" 7 | min_ngram: 1 8 | max_ngram: 1 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "char_wb" 11 | min_ngram: 1 12 | max_ngram: 5 13 | - name: "DIETClassifier" 14 | epochs: 200 15 | batch_size: [64, 128] 16 | entity_recognition: True 17 | BILOU_flag: Flase 18 | intent_classification: True 19 | use_masked_language_model: True 20 | transformer_size: 256 21 | number_of_transformer_layers: 2 22 | number_of_attention_heads: 4 23 | hidden_layers_sizes: 24 | text: [] 25 | label: [] 26 | dense_dim: 27 | text: 512 28 | label: 20 29 | scale_loss: False 30 | use_sparse_input_dropout: True 31 | weight_sparsity: 0 32 | 33 | data: 34 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 35 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 36 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-1/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-4/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: False 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-2/entity-recognition/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: False 19 | intent_classification: False 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-ConveRT-mask-loss/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "ConveRTTokenizer" 5 | - name: "ConveRTFeaturizer" 6 | - name: "CountVectorsFeaturizer" 7 | analyzer: "word" 8 | min_ngram: 1 9 | max_ngram: 1 10 | - name: "CountVectorsFeaturizer" 11 | analyzer: "char_wb" 12 | min_ngram: 1 13 | max_ngram: 5 14 | - name: "DIETClassifier" 15 | epochs: 200 16 | batch_size: [64, 128] 17 | entity_recognition: True 18 | BILOU_flag: Flase 19 | intent_classification: True 20 | use_masked_language_model: True 21 | transformer_size: 256 22 | number_of_transformer_layers: 2 23 | number_of_attention_heads: 4 24 | hidden_layers_sizes: 25 | text: [] 26 | label: [] 27 | dense_dim: 28 | text: 512 29 | label: 20 30 | scale_loss: False 31 | use_sparse_input_dropout: True 32 | weight_sparsity: 0 33 | 34 | data: 35 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 36 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 37 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: False 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-GloVe-mask-loss/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "SpacyNLP" 5 | - name: "SpacyTokenizer" 6 | - name: "SpacyFeaturizer" 7 | - name: "CountVectorsFeaturizer" 8 | analyzer: "word" 9 | min_ngram: 1 10 | max_ngram: 1 11 | - name: "CountVectorsFeaturizer" 12 | analyzer: "char_wb" 13 | min_ngram: 1 14 | max_ngram: 5 15 | - name: "DIETClassifier" 16 | epochs: 200 17 | batch_size: [64, 128] 18 | entity_recognition: True 19 | BILOU_flag: Flase 20 | intent_classification: True 21 | use_masked_language_model: True 22 | transformer_size: 256 23 | number_of_transformer_layers: 2 24 | number_of_attention_heads: 4 25 | hidden_layers_sizes: 26 | text: [] 27 | label: [] 28 | dense_dim: 29 | text: 512 30 | label: 20 31 | scale_loss: False 32 | use_sparse_input_dropout: True 33 | weight_sparsity: 0 34 | 35 | data: 36 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 37 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 38 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: False 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-1.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_1/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_1/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-2.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_2/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_2/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-3.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_3/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_3/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-4.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_4/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_4/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-5.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_5/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_5/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-6.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_6/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_6/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-7.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_7/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_7/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-8.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_8/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_8/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-9.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_9/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_9/test.md" 40 | -------------------------------------------------------------------------------- /configs/table-3/sparse-BERT-mask-loss/config-NLU-Evaluation-Data-Fold-10.yml: -------------------------------------------------------------------------------- 1 | language: en 2 | 3 | pipeline: 4 | - name: "HFTransformersNLP" 5 | model_name: "bert" 6 | model_weights: "bert-base-uncased" 7 | - name: "LanguageModelTokenizer" 8 | - name: "LanguageModelFeaturizer" 9 | - name: "CountVectorsFeaturizer" 10 | analyzer: "word" 11 | min_ngram: 1 12 | max_ngram: 1 13 | - name: "CountVectorsFeaturizer" 14 | analyzer: "char_wb" 15 | min_ngram: 1 16 | max_ngram: 5 17 | - name: "DIETClassifier" 18 | epochs: 200 19 | batch_size: [64, 128] 20 | entity_recognition: True 21 | BILOU_flag: Flase 22 | intent_classification: True 23 | use_masked_language_model: True 24 | transformer_size: 256 25 | number_of_transformer_layers: 2 26 | number_of_attention_heads: 4 27 | hidden_layers_sizes: 28 | text: [] 29 | label: [] 30 | dense_dim: 31 | text: 512 32 | label: 20 33 | scale_loss: False 34 | use_sparse_input_dropout: True 35 | weight_sparsity: 0 36 | 37 | data: 38 | train_file: "data/NLU-Evaluation-Data/KFold_10/train.md" 39 | test_file: "data/NLU-Evaluation-Data/KFold_10/test.md" 40 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # DIET: Lightweight Language Understanding for Dialogue Systems 2 | 3 | Source code to reproduce results of our [paper](https://arxiv.org/pdf/2004.09936.pdf) 4 | "DIET: Lightweight Language Understanding for Dialogue Systems". 5 | 6 | In order to reproduce the experiments results, execute the following steps: 7 | 8 | (1) We used Rasa for running the experiments. 9 | You first need to clone the repository, checkout the branch `diet-paper` and install Rasa. 10 | 11 | ```bash 12 | # clone the repository and checkout 'diet-paper' branch 13 | git clone https://github.com/RasaHQ/rasa 14 | cd rasa 15 | git checkout diet-paper 16 | # installation 17 | curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/get-poetry.py | python3 18 | make install-full 19 | ``` 20 | 21 | (2) Execute the `run.sh` script to train and test models on ATIS, SNIPS, and NLU Evaluation Data. 22 | You need to specify what results you want to reproduce by specifying the table from the paper. 23 | Make sure you are at the root directory of this repository before executing the script. 24 | Consider executing the experiments on a machine with a GPU to speed up the experiments. 25 | 26 | ```bash 27 | ./run.sh [table-1|table-2|table-3|table-4|table-5] 28 | ``` 29 | 30 | (3) The experiments results can be found in the folder `experiments`. 31 | 32 | To test other featurization or hyperparameters update the configuration files in the `configs` folder. 33 | Available featurization components and a list of available hyperparameters can be found 34 | [here](https://rasa.com/docs/rasa/nlu/components/). 35 | -------------------------------------------------------------------------------- /run.sh: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env bash 2 | 3 | parse_yaml() { 4 | local prefix=$2 5 | local s='[[:space:]]*' w='[a-zA-Z0-9_]*' fs=$(echo @|tr @ '\034') 6 | sed -ne "s|^\($s\)\($w\)$s:$s\"\(.*\)\"$s\$|\1$fs\2$fs\3|p" \ 7 | -e "s|^\($s\)\($w\)$s:$s\(.*\)$s\$|\1$fs\2$fs\3|p" $1 | 8 | awk -F$fs '{ 9 | indent = length($1)/2; 10 | vname[indent] = $2; 11 | for (i in vname) {if (i > indent) {delete vname[i]}} 12 | if (length($3) > 0) { 13 | vn=""; for (i=0; i "train.log" 53 | rasa test nlu --nlu "$CURRENT_DIR/$config_data_test_file" --config "$CURRENT_DIR/$filename" &> "test.log" 54 | 55 | python $CURRENT_DIR/evaluation_scripts/evaluation_nlu_evaluation_data.py -i results/diet-paper-eval.json 56 | python $CURRENT_DIR/evaluation_scripts/evaluation_atis_snips.py -i results/diet-paper-eval.json 57 | 58 | cd $CURRENT_DIR 59 | cp $filename $EXPERIMENT_FOLDER 60 | 61 | CURRENT_EXPERIMENT=$((CURRENT_EXPERIMENT + 1)) 62 | done 63 | 64 | # calculate avg results for NLU Evaluation Data 65 | 66 | CHECKED_FOLDERS=() 67 | 68 | for filename in $FILES; do 69 | NAME=$(basename "$filename" .yml) 70 | PARENTDIR="$(basename "$(dirname "$filename")")" 71 | 72 | if [[ "$PARENTDIR" = "$TABLE_FOLDER" ]]; then 73 | EXPERIMENT_FOLDER=experiments/$TABLE_FOLDER/ 74 | else 75 | EXPERIMENT_FOLDER=experiments/$TABLE_FOLDER/$PARENTDIR/ 76 | fi 77 | 78 | if [[ "$NAME" == *"NLU-Evaluation-Data"* ]]; then 79 | if ! [[ $CHECKED_FOLDERS =~ (^|[[:space:]])$EXPERIMENT_FOLDER($|[[:space:]]) ]]; then 80 | python evaluation_scripts/evaluation_nlu_evaluation_data.py -f "$EXPERIMENT_FOLDER/config-NLU-Evaluation-Data-Fold-{}" 81 | fi 82 | fi 83 | 84 | CHECKED_FOLDERS+=($EXPERIMENT_FOLDER) 85 | done -------------------------------------------------------------------------------- /evaluation_scripts/converters.py: -------------------------------------------------------------------------------- 1 | import os 2 | import json 3 | 4 | 5 | # Code taken from 6 | # https://gitlab.com/hwu-ilab/hermit-nlu/blob/master/data/nlu_benchmark/ 7 | # to ensure same evaluation metrics as https://arxiv.org/abs/1910.00912 8 | # Additional methods were added to be able to call desired evaluation methods. 9 | 10 | 11 | arg_format_pattern = r"\[\s*(?P