├── README.md
└── facebook_gaph_sentiment_analysis.py


/README.md:
--------------------------------------------------------------------------------
1 | # sentiment_analysis_graph
2 | a sentiment analysis of comments in a facebook page using python 3.5 facebook graph api and NLTK 
3 | 


--------------------------------------------------------------------------------
/facebook_gaph_sentiment_analysis.py:
--------------------------------------------------------------------------------
  1 | #translator to translate comments from others languages to english
  2 | from  translate import Translator
  3 | 
  4 | import requests
  5 | import signal
  6 | import sys
  7 | import pprint
  8 | graph_api_version = 'v2.9'
  9 | 
 10 | # add your token instead
 11 | access_token = access_token
 12 | translator= Translator(to_lang="fr")
 13 | 
 14 | # the user id for me the facebook page of GOAL  ENGLISH
 15 | user_id = '25427813597'
 16 | limit = 200
 17 | posts = []
 18 | url_profile = 'https://graph.facebook.com/{}/{}?fields=feed'.format(graph_api_version,user_id)
 19 | 
 20 | r_profile = requests.get(url_profile, params={'access_token': access_token})
 21 | while True:
 22 |     data = r_profile.json()
 23 |     feed = data['feed']
 24 |     for post in feed['data']:
 25 |         posts.append(post['id'])
 26 |     if 0 < limit <= len(posts):
 27 |         break
 28 |     if 'paging' in data and 'next' in data['paging']:
 29 |         r_profile = requests.get(data['paging']['next'])
 30 |     else:
 31 |         break
 32 | 
 33 | comments = []
 34 | 
 35 | 
 36 | post_index = 0
 37 | for post in posts :
 38 |     url = 'https://graph.facebook.com/{}/{}/comments'.format(graph_api_version,post)
 39 |     r = requests.get(url, params={'access_token': access_token})
 40 |     print('post Number :',post_index=+1)
 41 |     while True:
 42 |         data = r.json()
 43 | 
 44 |         # catch errors returned by the Graph API
 45 |         if 'error' in data:
 46 |             raise Exception(data['error']['message'])
 47 | 
 48 |         # append the text of each comment into the comments list
 49 |         for comment in data['data']:
 50 |             # remove line breaks in each comment
 51 |             text = comment['message'].replace('\n', ' ')
 52 | 
 53 |             #translate the comments to english before add them
 54 |             text = translator.translate(text)
 55 |             comments.append(text)
 56 | 
 57 |         print('Got {} comments, total: {}'.format(len(data['data']), len(comments)))
 58 | 
 59 |         # check if we have enough comments
 60 |         if 0 < limit <= len(comments):
 61 |             break
 62 | 
 63 |         # check if there are more comments
 64 |         if 'paging' in data and 'next' in data['paging']:
 65 |             r = requests.get(data['paging']['next'])
 66 |         else:
 67 |             break
 68 | 
 69 | 
 70 | #sentiment analysis part
 71 | from nltk.sentiment  import vader
 72 | sid = vader.SentimentIntensityAnalyzer()
 73 | 
 74 | output = [{ 'sentence' :sentence , 'scores' : sid.polarity_scores(sentence) } for sentence in comments ]
 75 | import json
 76 | sentiment_analysis = open('sentiment_analysis_'+user_id+'_en.json','w')
 77 | json_out = json.dump(output,sentiment_analysis)
 78 | pprint.pprint(json_out)
 79 | 
 80 | import pandas as pd
 81 | import numpy as np
 82 | df = pd.read_json('sentiment_analysis_'+user_id+'_en.json').fillna(' ')
 83 | 
 84 | # the mean of sentiment in a dataframe contains comments
 85 | def mean_senti(df) :
 86 |     neg = []
 87 |     pos = []
 88 |     neutre = []
 89 |     for e in df["scores"] :
 90 |         neg.append(e["neg"])
 91 |         pos.append(e["pos"])
 92 |         neutre.append(e["neu"])
 93 |     neg = np.array(neg)
 94 |     pos = np.array(pos)
 95 |     neut = np.array(neutre)
 96 |     neutre = np.mean(neut, axis=0)
 97 |     positive = np.mean(pos, axis=0)
 98 |     negative = np.mean(neg, axis=0)
 99 |     return(negative,positive,neutre)
100 | 
101 | #make a sub dataframe of comments contains the term "zidan"
102 | sub_def = df[df['sentence'].str.contains("zidan")]
103 | 
104 | (neg_c,pos_c,neutre_c) = mean_senti(sub_def)
105 | print('negative :',neg_c,'positive :',pos_c,'neutre :',neutre_c)
106 | 


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