-
Notifications
You must be signed in to change notification settings - Fork 0
/
sentiment.py
175 lines (130 loc) · 4.37 KB
/
sentiment.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
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
import nltk
import random
import pickle
from nltk.classify.scikitlearn import SklearnClassifier
from sklearn.naive_bayes import MultinomialNB, BernoulliNB
from sklearn.linear_model import LogisticRegression, SGDClassifier
from sklearn.svm import SVC, LinearSVC, NuSVC
from nltk.classify import ClassifierI
from statistics import mode
from nltk.tokenize import word_tokenize
from election import NLPElection
class Sentiment:
'''
General purpose sentiment analyser. It relies on pickled
training and testing sets as well as pretrained classifiers
to work.
'''
def __init__(self, path):
word_features5k_f = open(path, "rb")
self.word_features = pickle.load(word_features5k_f)
word_features5k_f.close()
def train(self, nltk, mnb, bnb, lgr, svc):
self.nltk_classifier_nb(nltk)
self.multinomial_nb(mnb)
self.bernoulli_nb(bnb)
self.logistic_regress(lgr)
self.linear_svc(svc)
self.election()
def find_features(self, document):
words = word_tokenize(document)
features = {}
for word in self.word_features:
features[word] = (word in words)
return features
def create_sets(self, path, boundary=60):
'''
This method is for creating training and testing
sets for your classifiers. It actually has no use
for this module, but serves as a reference
'''
featuresets_f = open(path, "rb")
self.feature_sets = pickle.load(featuresets_f)
featuresets_f.close()
random.shuffle(self.feature_sets)
length = len(self.feature_sets)
limit = length * boundary / 100
self.training_set = feature_sets[:limit]
self.testing_set = feature_sets[limit:]
def nltk_classifier_nb(self, path):
open_file = open(path, "rb")
self.classifier = pickle.load(open_file)
open_file.close()
def multinomial_nb(self, path):
open_file = open(path, "rb")
self.MNB_classifier = pickle.load(open_file)
open_file.close()
def bernoulli_nb(self, path):
open_file = open(path, "rb")
self.BernoulliNB_classifier = pickle.load(open_file)
open_file.close()
def logistic_regress(self, path):
open_file = open(path, "rb")
self.LogisticRegression_classifier = pickle.load(open_file)
open_file.close()
def linear_svc(self, path):
open_file = open(path, "rb")
self.LinearSVC = pickle.load(open_file)
open_file.close()
def election(self):
self.election_classifier = NLPElection(
self.classifier,
self.MNB_classifier,
self.BernoulliNB_classifier,
self.LogisticRegression_classifier,
self.LinearSVC
)
def analyse(self, data, lower_limit):
'''
Performs the sentiment analysis for a list of text data
'''
self.sentiments = []
for text in data:
features = self.find_features(text)
if self.election_classifier.unanimity(features) < lower_limit:
self.sentiments.append(self.BernoulliNB_classifier.classify(features))
else:
self.sentiments.append(self.election_classifier.vote(features))
def get_sentiment(self, data, lower_limit=0.4):
'''
Used for deriving sentiment on a list of data around
a particular topic, such as stock trends or tweets
'''
self.analyse(data, lower_limit)
total = len(self.sentiments)
proportion = self.sentiments.count('pos') / total
if proportion > 0.75:
return 'very good'
elif proportion > 0.5:
return 'good'
elif proportion > 0.25:
return 'bad'
else:
return 'very bad'
def get_sentiment_single(self, text):
'''
Used for deriving sentiment on a single body of text,
such as a review or critique
'''
features = self.find_features(text)
if self.election_classifier.unanimity(features) > 0.75:
if self.election_classifier.vote(features) == 'pos':
return 'very good'
else:
return 'very bad'
else:
if self.election_classifier.vote(features) == 'pos':
return 'good'
else:
return 'bad'
class TwitterSentiment(Sentiment):
def __init__(self):
word_features5k_f = open("pickled/twitter/resources/word_features5k.pickle", "rb")
self.word_features = pickle.load(word_features5k_f)
word_features5k_f.close()
self.nltk_classifier_nb('pickled/twitter/classifiers/originalnaivebayes5k.pickle')
self.multinomial_nb('pickled/twitter/classifiers/MNB_classifier5k.pickle')
self.bernoulli_nb('pickled/twitter/classifiers/BernoulliNB_classifier5k.pickle')
self.logistic_regress('pickled/twitter/classifiers/LogisticRegression_classifier5k.pickle')
self.linear_svc('pickled/twitter/classifiers/LinearSVC_classifier5k.pickle')
self.election()