/sentiment-analysis

Sentiment Analysis Algorithms for tweets

Primary LanguagePython

Tweets Sentiment Analysis

Sentiment analysis on Twitter data has attracted much attention recently. This program is a simple implementation of a sentiment classifier for a tweet.

Based on the sentiwordnet corpus, the classifier can distinguish if a tweet is Positive, Negative, or Objective.

Usage

  • To see answers to the TP : python tp.py
  • To see the results of the classification using the first version of the algorithm : python tpv1.py
  • To see the results of the classification using the second version of the algorithm : python tpv2.py
    • To see the results of the classification using the third version of the algorithm : python tpv3.py

Requirements

You need the following libs :

  • sklearn
  • nltk
  • numpy

Processing pipeline

The processing pipeline for every tweet is the following :

Algorithm v1

  • Tweet preprocessing : removal of twitter-specific characters ( mentions : @, hashtags : #, retweets : RT)
  • POS tagging : assign every word to its POS tag (Part Of Speech Tag)
  • SentiSynset : For words such as adjective, words, nouns, and adverb, find the first sentisynset in the sentiwordnet corpus
  • Assign a score for the tweet by aggregating every individual score from the sentisynsets
  • Classify the tweet

Algorithm v2 : taking negation and modifiers into account

  • Tweet preprocessing : removal of twitter-specific characters ( mentions : @, hashtags : #, retweets : RT)
  • POS tagging : assign every word to its POS tag (Part Of Speech Tag)
  • SentiSynset : For words such as adjective, words, nouns, and adverb, find the first sentisynset in the sentiwordnet corpus
  • Assign a score for the tweet by aggregating every individual score from the sentisynsets
  • Modify the score to take care of negations and modifiers
  • Classify the tweet

Algorithm v3 : taking emojis into account

  • Tweet preprocessing : removal of twitter-specific characters ( mentions : @, hashtags : #, retweets : RT)
  • POS tagging : assign every word to its POS tag (Part Of Speech Tag)
  • SentiSynset : For words such as adjective, words, nouns, and adverb, find the first sentisynset in the sentiwordnet corpus
  • Assign a score for the tweet by aggregating every individual score from the sentisynsets
  • Modify the score to take care of negations and modifiers
  • Increase the positive or negative score by taking care of emojis
  • Classify the tweet

Classify tweets

python tpv1.py

python tpv2.py

python tpv3.py