/
semantic.py
157 lines (116 loc) · 6.33 KB
/
semantic.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Thu Aug 29 09:11:19 2019
@author: sanjanamendu
"""
from tqdm import tqdm
import pandas as pd
import string
import sklearn
import sklearn.decomposition
import contractions
import itertools
import nltk
import re
import os
tqdm.pandas()
home = os.path.expanduser("~")
mallet_path = home+"/Downloads/Mallet/"
# =============================================================================
# TF-IDF
# =============================================================================
fb_agg = pd.read_csv(mallet_path+"fbmsg_agg.csv").drop('Unnamed: 0',1)
fb_agg.Clean_Content = fb_agg.Clean_Content.fillna('')
fb_agg.Clean_Content = fb_agg.groupby('PID').Clean_Content.transform(lambda x: ' '.join(x))
fb_agg = fb_agg.groupby('PID').first().reset_index()
fb_agg.Clean_Content = fb_agg.Clean_Content.str.replace('\s\s+','',regex=True)
word_vectorizer = sklearn.feature_extraction.text.TfidfVectorizer(ngram_range=(1,3), analyzer='word')
X = word_vectorizer.fit_transform(fb_agg.Clean_Content)
tfidf = pd.DataFrame(X.toarray(), columns=word_vectorizer.get_feature_names(), index=fb_agg.PID)
word_vectorizer = sklearn.feature_extraction.text.CountVectorizer(ngram_range=(1,3), analyzer='word')
X = word_vectorizer.fit_transform(fb_agg.Clean_Content)
frequencies = pd.DataFrame(X.toarray(), columns=word_vectorizer.get_feature_names(), index=fb_agg.PID)
binary = (frequencies > 0).astype(int)
min_participants = 10
vocab = [x for x in binary.columns.values[binary.sum(axis=0) > min_participants] \
if not x.replace(' ','').isdigit()]
# Pointwise Mutual Information
text = " ".join(fb_agg.Clean_Content)
trigram_measures = nltk.collocations.TrigramAssocMeasures()
trigram_coloc = nltk.collocations.TrigramCollocationFinder.from_words(nltk.tokenize.word_tokenize(text))
trigram_pmi = trigram_coloc.score_ngrams(trigram_measures.pmi)
trigram_keep = [' '.join(term) for term, pmi in trigram_pmi if pmi > 9]
bigram_measures = nltk.collocations.BigramAssocMeasures()
bigram_coloc = nltk.collocations.BigramCollocationFinder.from_words(nltk.tokenize.word_tokenize(text))
bigram_pmi = bigram_coloc.score_ngrams(bigram_measures.pmi)
bigram_keep = [' '.join(term) for term, pmi in bigram_pmi if pmi > 6]
final_vocab = [v for v in tqdm(vocab) if len(v.split()) == 1 or v in bigram_keep or v in trigram_keep]
tfidf[final_vocab].to_csv("fbmsg_term_tfidf.csv")
#binary[final_vocab].to_csv("fbmsg_term_binary.csv")
#
#final_freq = frequencies[final_vocab]
#final_freq.to_csv("fbmsg_term_freq.csv")
#
#final_relativfreq = final_freq.div(final_freq.sum(axis=1), axis=0) # normalize
#final_relativfreq.to_csv("fbmsg_term_relativfreq.csv")
#
#final_anscombe = final_freq.apply(lambda x: 2 * np.sqrt(x + 0.375)) # apply Anscombe transformation
#final_anscombe.to_csv("fbmsg_term_anscombe.csv")
# =============================================================================
# TOPIC MODELING
# =============================================================================
fb_agg = pd.read_csv("fbmsg_agg (participant).csv")
fb_agg.Content = fb_agg.Content.fillna('')
# Lists of Words to Omit
stop = nltk.corpus.stopwords.words('english')
extra_stop = ['get','got','tho','one',"i'll","i'm","that's","can't",'yall','hey',"i've",'also',\
'dont','thats','yes','ani','tah','thats','yea','dont','hey','also',"i'll",'that!',\
'ur',"It's","i've","i'm","'cause","they'll","i'll","that's","how'd",\
"who's","let's",'ill','want','also','might',"he's",'cant',"y'all",'hey','much',\
'good','also','doesnt','didnt','wont',"we're",'whats',"there's","what's","ain't",\
'imma','kekek','idek','maybe','itll','isnt','that ','hows','would',"there's",\
'havent','zach','kkkk',"what's","we'll",'isnt','whats','maybe','u','im']
friends_names = ['katie','erik','chris','johnny','alan','brandon','andrew','daniel','tony',\
'richie','adrien','john','andrew','stephen','crystal','david','juhan','erik',\
'edmund','claire','sadia','jenny','chelsea','tony','kirtana','kadariya','bishal',\
'tony','nihar','adrien','tina','erica','pooja','hajur','steph','rohan','kuo','rashid']
names = [n.lower() for n in nltk.corpus.names.words()]
punctuation = list(string.punctuation)
# Concatenate all lists of words to omit (except punctuation)
no_punc = stop + extra_stop + friends_names + names
# Regex + Non-Regex List of Words to Omit
rm_words = no_punc + punctuation
rm_regex = no_punc + [re.escape(x) for x in punctuation]
# Vocabulary
tknzr = nltk.tokenize.TweetTokenizer(preserve_case=False, strip_handles=True, reduce_len=True)
pat = r'\b(?:{})\b'.format('|'.join(rm_regex))
tokenized = fb_agg.Content.progress_apply(lambda x: ' '.join([t for t in tknzr.tokenize(x) if t not in rm_words]))
fb_agg['Tokenized_Content'] = tokenized
def clean_text(x):
x = ' '.join([t for t in tknzr.tokenize(x)])
x = contractions.fix(x) # Expand Contractions
x = ''.join(''.join(s)[:2] for _, s in itertools.groupby(x)) # Standardize
x = re.sub('http\S+','',x) # Remove URLs
x = re.sub('www\S+','',x) # Remove URLs
x = re.compile(pat, re.I).sub("",x) # Remove Stopwords
x = re.sub(r'\b\w{1,2}\b','',x) # Remove words with < 2 letters
x = re.sub('\s\s+',' ',x) # Remove Extra Spaces
return x
fb_agg['Clean_Content'] = fb_agg.Tokenized_Content.progress_apply(lambda x: clean_text(x))
fb_agg = fb_agg.replace(r'\\n',' ', regex=True)
fb_agg[["PID","Clean_Content"]].to_csv("fbmsg_agg.csv")
# --- Mallet Commands
os.system(mallet_path + "bin/mallet import-file " + \
"--input "+ mallet_path +"fbmsg_agg.csv --output fbmsg_agg.mallet " + \
"--keep-sequence")
os.system(mallet_path+"bin/mallet train-topics --config "+mallet_path+"config.txt")
doc_topic = pd.read_csv("fbmsg_agg_doc-topics.csv", header=None)
doc_topic.columns = ['Document'] + ['Topic '+str(i) for i in range(100)]
doc_topic["PID"] = doc_topic.pop('Document').str.split(',',expand=True).iloc[:,1]
doc_topic = doc_topic[doc_topic.PID.str.len() == 4]
avg_doc_topic = doc_topic.groupby('PID').mean().reset_index()
avg_doc_topic.to_csv("fbmsg_doc_topic.csv",index=False)
topic_usage = pd.read_csv("fbmsg_doc_topic.csv")
tfidf = pd.read_csv("fbmsg_term_tfidf.csv")
semantic_feat = pd.merge(tfidf, avg_doc_topic, 'left', on='PID')