FVEC feature and Machine Learning Approach for Indonesian Opinion Mining on YouTube Comments

Abstract

Mining opinions from Indonesian comments from YouTube videos are required to extract interesting patterns and valuable information from consumer feedback. Opinions can consist of a combination of sentiments and topics from comments. The features considered in the mining of opinion become one of the important keys to getting a quality opinion. This paper proposes to utilize FVEC and TF-IDF features to represent the comments. In addition, two popular machine learning approaches in the field of opinion mining, i.e., SVM and CNN, are explored separately to extract opinions in Indonesian comments of YouTube videos. The experimental results show that the use of FVEC features on SVM and CNN achieves a very significant effect on the quality of opinions obtained, in term of accuracy.