A Comparative Analysis of Lexical/NLP Method with WEKA's Bayes Classifier

Abstract

Various websites are available as source of microblogs. This is due to nature of microblogs on which people post real time messages about their attitudes on a various topics, talk about present issues, criticize, and articulate positive or negative sentiment for products they use in daily life. That?s why, manufacturing companies of such products have started to take these microblogs to get a sense of general sentiment for their product. Reply can be given by the companies on microblogs for the reactions of the users. Thus challenge is to build a technique to detect and summarize an overall sentiment. The proposed methodology examines sentiments on Twitter data contextually. Sentiment Analysis is the major aspect of present day NLP. Also, Twitter has emerged as the most important data source for present day NLP. In the work carried out, tweets are extracted from Twitter using Twitter API after authentication, a fine pre-processing is dealt and provided for further processing. Later, tag each word with their respective parts of speech using Part-Of-Speech (POS) tagger. SentiWordNet, WordNet and NLP weight assignment policies are used to assign weights and provide results. The analysis of same data set is also done with Na?ve Bayes classifier using WEKA - the data mining tool. Then results of both ? the proposed method and Na?ve Bayes are compared. (Then finally comparison between the results of proposed method with Na?ve Bayes classier is done.) The investigation proved that our method i.e. NLP technique works better than that of Na?ve Bayes Classifier. And this study also proves that the training set to the classier matters a lot in Machine Learning - ?Expected output can be accurate if and only if the training of a classifier is better?.