class LexiconSentimentAnalyzer(object):
NEUTRAL_S_LIMIT = (-0.05, 0.05)
def __init__(self):
self.lex_reader = LexiconReader()
self.vader = VaderAnalyzer()
def calculate_simple_sentiment(self, s_dictionary_name, input_text):
"""
Calculate text sentiment in a simple way - sum of sentiments of all words in the text.
Arguments:
s_dictionary_name (string)
input_text (string)
Returns:
float: sentiment value
"""
# Get dictionary
s_dict = self.lex_reader.get_dictionary(s_dictionary_name)
# Tokenize text
tokens = nltk.word_tokenize(input_text.lower())
# Remove punctation
tokens = [i for i in tokens if i not in string.punctuation]
# Calculate sentiment
sentiment_sum = 0
found_tokens_count = 0.0
for token in tokens:
if token in s_dict:
found_tokens_count += 1
sentiment_sum += s_dict[token]
# count percentage of found tokens
percent_tokens_found = (found_tokens_count / len(tokens)) * 100
# result
return round(sentiment_sum, 3), round(percent_tokens_found, 3)
def calculate_vader_sentiment(self, s_dictionary_name, input_text, split_sentences=False):
"""
Calculate text sentiment using VADER algorithm and selected dictionary.
Arguments:
s_dictionary_name (string)
input_text (string)
Returns:
float: sentiment value
"""
# If necessary, reload VADER object.
if s_dictionary_name != self.vader.lexicon_name:
self.vader = VaderAnalyzer(s_dictionary_name)
self.vader.lexicon_name = s_dictionary_name
# test
#print('reload')
#print len(self.vader.lexicon.keys())
#print self.vader.lexicon['}:-)']
# Calculate sentiment for the whole text.
if not split_sentences:
return self.vader.polarity_scores(input_text)['compound']
# If desired, split text into sentences.
sentences = nltk.tokenize.sent_tokenize(input_text)
# Calc sentiment for every sentence.
sentiment_sum = 0.0
for sent in sentences:
values = self.vader.polarity_scores(sent)
sentiment_sum += values['compound']
# Calc sentiment for the whole text - normalize sum by number of sentences.
return sentiment_sum / float(len(sentences))
def format_sentiment_value(self, sent_value):
"""
Return whether given sentiment value represents a positive, negative or neutral sentiment.
:param sent_value (float):
:return: string
"""
if self.NEUTRAL_S_LIMIT[0] < sent_value < self.NEUTRAL_S_LIMIT[1]:
polarity = 'neu'
elif sent_value > 0:
polarity = 'pos'
elif sent_value < 0:
polarity = 'neg'
else:
polarity = 'neu'
return polarity
def calculate_vader_sentiment_values(self, s_dictionary_name, input_text):
"""
Calculate text sentiment using VADER algorithm and selected dictionary.
Return compund value in form of sum and division.
Arguments:
s_dictionary_name (string)
input_text (string)
Returns:
floats: compound sum, compound division
"""
# If necessary, reload VADER object.
if s_dictionary_name != self.vader.lexicon_name:
self.vader = VaderAnalyzer(s_dictionary_name)
self.vader.lexicon_name = s_dictionary_name
# Split text into sentences.
sentences = nltk.tokenize.sent_tokenize(input_text)
# Calc sentiment for every sentence.
sentiment_sum = 0.0
for sent in sentences:
values = self.vader.polarity_scores(sent)
sentiment_sum += values['compound']
# Calc sentiment for the whole text - normalize sum by number of sentences.
sentiment_division = float(sentiment_sum / len(sentences))
# result
return sentiment_sum, round(sentiment_division, 4)
def __init__(self):
self.lex_reader = LexiconReader()
self.vader = VaderAnalyzer()