Twitter Sentiment Analysis Using NLTK and Machine Learning

Abstract

Abstract: In today’s online social networks like twitter all people choose to express their opinions on social networking sites about the products or organizations, if any of the user has a good experience with any of the product or company, he/she will express their views which can be good reviews/opinion by seeing these opinion other users can know the quality of the product. On Internet, opinion mining which can be on sentiments or topic helps users to know the quality of any of the organizations or products, while developing new techniques to detect the sentiments from these opinions, all existing techniques that are used to discover are either Positive or Negative or Neutral sentiments from topics but this paper proposes 5 levels of sentiments detection such as High Positive, Moderate Positive, Neutral, High Negative and Moderate Negative. To detect sentiments, we are using four Ordinal Regression machine learning algorithms such as SoftMax, Decision Tree, Random Forest and also Support Vector Regression. For classification of tweets, we used NLTK, which cleans the tweets by removing special symbols, removing stop words, word stemming, etc. In this paper the authors have discussed how these algorithms are implemented on tweets and detect the sentiments