Preprocessing of YouTube Myanmar Music Comments for Sentiment Analysis

Abstract

Over the years, social media sites have become an important role of communication and a source of the tremendous amount of real-time data of images, videos, etc. YouTube is likely the most popular of them, with millions of uploaded videos and billions of comments for all these videos. This paper presents a new Music dataset, which has the YouTube comments written in Myanmar language, to be applied in sentiment analysis. Data preprocessing is very important for our language because it allows improving the quality of the raw data and converting the raw data into a clean dataset. The preprocessing of music comments is followed by basic phase, removing phase, segmentation phase, replacement phase and translation phase. The outcome of YouTube comment preprocessing will aid in better sentiment analysis. Results show that the preprocessing approaches give a significant effect on the musical opinion extraction process using information gain.