Building Machine Learning Based Senti-word Lexicon for Sentiment Analysis

Sentiment analysis involves classifying opinions in text into categories like "positive" or "negative". One of approaches used to make sentiment classification is using sentiment lexicon. This paper aims to build a sentiment lexicon which is domain independent. We propose a Machine Learning Based Senti-word Lexicon (MLBSL) based on the Amazon data set which contains reviews from different domains. Our proposed MLBSL yields an improvement over previous published manual and automatic-built lexicons like SentiWordNet. We also provide an improvement in calculation method used in reviews sentiment analysis.