Experiment with Lexicon Based Techniques on Domain-Specific Malay Document Sentiment Analysis

Abstract

The nature of Sentiment Analysis (SA) mostly is generated by human beings. They expressed their emotion in writing or expressing their feeling via social media or blog. The Advancement of Internet and the increasing number of users in social media are the factors on why the sentiment analysis gaining its popularity in Malay languages. This research aims to implement the Sentiment Analysis on Malay language documents and propose a lexicon-based technique for Malay based sentiment analysis on specific domain such as Song, Politic and Product to find the best SA classifier on the Domain-Specific Malay Document Sentiment Analysis. Analysis of the evaluation result is based on the comparison of expert evaluation, Lexicon-based evaluation’s result and Naïve Bayes SA classification’s result, which is Naïve Bayes represent Machine Learning approach in this study. The result shows Lexicon-based Classification has outperformed Naïve Bayes SA classification in overall 3 topics which are Song, Politic and Product in average of 70% compared to 50% average for Naïve Bayes. For the future works, the researcher would like to improve in the particular area such as Sentiment Analysis based on the Malay dialect, increase the data in the dictionary and applying phrase level for better and optimum results.