An Improved Sentiment Classification using Lexicon into SVM

Abstract

With the emergence of web 2.0 and availability of huge amount of digital data on the social web, people always want to discover unknown, to predict events that could occur, and the procedure on how it works and change over time. Similarly, sentiment analysis is related with the automatic extraction of sentiment information from textual data available at various social webs. While most sentiment analysis deals commercial jobs like fetching opinions from product reviews, there is significant growth in social web and it becomes a source to promote various products. This is actual reason why most of the commercial web support login through social web like facebook, twitter. There are two approaches to sentiment analysis. First one is based on lexicon and second is machine learning. It has been proved that machine learning approach performs better than lexicon based approaches but it ignores the knowledge encoded in sentiment lexicons. This paper describes a method that includes sentiment lexicons as prior information to SVM approach, a machine learning techniques, to improve the accuracy of sentiment analysis. It also describes a technique to automatically create domain specific sentiment lexicons for this learning purpose. A result shows that the domain specific lexicons lead to a significant accuracy improvement for sentiment analysis.