Paper Title
Sentiment Orientation System of Automotive Reviews using Multinomial Naive Bayes Classifier at Document Level
Abstract
The abundance of opinions available on the World Wide Web represents an information repository of enormous
intellectual and economic value. Web is used in every field. Almost of all people use web for an ordinary purpose like online
shopping sites, blogs, social network sites, forums etc. The large numbers of reviews are given by the users that reflect
whether the product is good or bad. People’s opinions and experience are very valuable information in decision making
process. Opinion Mining or Sentiment Analysis is a natural language processing task that concerns with finding orientation
of opinion in a piece of text with respect to a topic. This paper focuses on document level opinion mining and proposes
clustering documents according to their polarity scores using K-means algorithm and classifying the particular documents
based on the clusters using Naïve Bayes classifier. In the proposed system, TF-IDF approach is applied to denote how
important a word is to a document and SentiWordNet supports the system to determine the scores of opinion words. The
experimental work was done on automotive reviews. The proposed method achieved total accuracy of 93.7% on the test set.
Index terms - K-means, Naïve Bayes classifier, Opinion mining, Sentiment analysis, SentiWordNet, TF-IDF approach.