├── README.md
└── Subject_Verb_Object_Extractor.py


/README.md:
--------------------------------------------------------------------------------
1 | # Extract-SVO
2 | A python script to extract subject-predicate-object (SVO) triplets from English sentences using Stanford Parser according to the following paper "Rusu, Delia, et al. "Triplet extraction from sentences." Proceedings of the 10th International Multiconference" Information Society-IS. 2007."
3 | You can download the paper from the following link https://www.researchgate.net/profile/Bla_Fortuna/publication/228905420_Triplet_extraction_from_sentences/links/0912f50866261e6a54000000.pdf
4 | 
5 | Stanford Parser is a natural language parser developed by Dan Klein and Christopher D. Manning from The Stanford Natural Language Processing Group. you can download the package from https://stanfordnlp.github.io/CoreNLP/download.html 
6 | 
7 | Stanford Parser generates a Treebank parse tree for the input sentence.An example of a  parse tree for the sentence “A rare black squirrel has become a regular visitor to a suburban garden”. The triplet extracted out of this sentence is squirrel – become – visitor.
8 | 


--------------------------------------------------------------------------------
/Subject_Verb_Object_Extractor.py:
--------------------------------------------------------------------------------
 1 | import os
 2 | #Set standford parser and models in your environment variables.
 3 | os.environ['STANFORD_PARSER'] = 'stanford-parser'
 4 | os.environ['STANFORD_MODELS'] = 'stanford-parser'
 5 | from nltk.parse.stanford import StanfordParser
 6 | from nltk.tree import ParentedTree, Tree
 7 | 
 8 | parser = StanfordParser()
 9 | 
10 | def find_subject(t):    
11 |     for s in t.subtrees(lambda t: t.label() == 'NP'):
12 |         for n in s.subtrees(lambda n: n.label().startswith('NN')):
13 |             return (n[0], find_attrs(n))
14 |             
15 | def find_predicate(t):    
16 |     v = None
17 |     
18 |     for s in t.subtrees(lambda t: t.label() == 'VP'):
19 |         for n in s.subtrees(lambda n: n.label().startswith('VB')):
20 |             v = n
21 |         return (v[0], find_attrs(v))
22 |     
23 | def find_object(t):    
24 |     for s in t.subtrees(lambda t: t.label() == 'VP'):
25 |         for n in s.subtrees(lambda n: n.label() in ['NP', 'PP', 'ADJP']):
26 |             if n.label() in ['NP', 'PP']:
27 |                 for c in n.subtrees(lambda c: c.label().startswith('NN')):
28 |                     return (c[0], find_attrs(c))
29 |             else:
30 |                 for c in n.subtrees(lambda c: c.label().startswith('JJ')):
31 |                     return (c[0], find_attrs(c))
32 |                 
33 | def find_attrs(node):
34 |     attrs = []
35 |     p = node.parent()
36 |     
37 |     # Search siblings
38 |     if node.label().startswith('JJ'):
39 |         for s in p:
40 |             if s.label() == 'RB':
41 |                 attrs.append(s[0])
42 |                 
43 |     elif node.label().startswith('NN'):
44 |         for s in p:
45 |             if s.label() in ['DT','PRP$','POS','JJ','CD','ADJP','QP','NP']:
46 |                 attrs.append(' '.join(s.flatten()))
47 |     
48 |     elif node.label().startswith('VB'):
49 |         for s in p:
50 |             if s.label() == 'ADVP':
51 |                 attrs.append(' '.join(s.flatten()))
52 |                 
53 |     # Search uncles
54 |     if node.label().startswith('JJ') or node.label().startswith('NN'):
55 |         for s in p.parent():
56 |             if s != p and s.label() == 'PP':
57 |                 attrs.append(' '.join(s.flatten()))
58 |                 
59 |     elif node.label().startswith('VB'):
60 |         for s in p.parent():
61 |             if s != p and s.label().startswith('VB'):
62 |                 attrs.append(s[0])
63 |                 
64 |     return attrs
65 | 
66 | def main(sentence):
67 |     print find_subject(sentence)
68 |     print find_predicate(sentence)
69 |     print find_object(sentence)
70 | 
71 | if __name__=="__main__" :
72 |     import sys   
73 |     # Parse the example sentence
74 |     sent = 'A rare black squirrel has become a regular visitor to a suburban garden'
75 |     main(sent)
76 | 


--------------------------------------------------------------------------------