Sentiment Analysis

Abstract

Sentiment analysis is the technique to analyze people's perspective towards different entities. There are three main kinds of Sentiment Analysis. This paper focuses on Aspect based Sentiment Analysis. In Aspect based Sentiment Analysis, there are three main tasks to be done. First task is the Aspect Extraction. Second task is to extract Sentiment word for that particular Aspect. And third task is to identify sentiment polarity towards that particular Aspect.(i.e. whether the sentiment is positive or negative towards that Aspect.

This report implements two topic models SAS-model (Seeded Aspect and Sentiment Model) and ME-SAS (Maximum Entropy SAS) for Aspect extraction and Sentiment word extraction. ME-SAS is the extended model of SAS. So for base implementation, First SAS model is implemented and then ME-SAS model is implemented using Stanford-pos Tagger. So, First and Second tasks of Sentiment Analysis is performed by this models. After extracting Aspects and Aspect based Sentiment Word, to identify sentiment polarity, SentiWordNet, Lexical resource for opinion mining is used. To identify sentiment polarity, two methods SWN(AAC) that is SentiWordNet(Adjective+Adverb combine) and SWN(AAAVC) that is SentiWordNet(Adjective+Adverb and Adverb+verb) combine are used.