A COMPARATIVE STUDY OF GENRE-BASED SENTIMENT ANALYSIS

Abstract: The emergence of multiple online platforms has generated a huge volume of user-generated movie reviews from various sources like IMDb, Kaggle repositories, and Twitter. Due to the unstructured nature and diversity in using language, manual analysis is hard to conduct. Sentiment analysis hence serves as an effective means of automatic opinion polarity detection from textual data. However, sentiment expression in movie reviews is pretty much influenced by genre and dataset characteristics. This paper hence represents a comparison study of genre-oriented sentiment analysis based on both the lexicon-based approach and the machine learning-based approach. In this context, sentiment classification tasks are performed on a number of movie genres such as Naïve Bayes, Support Vector Machine, and Random Forest. The performances of these approaches are compared across a number of datasets, and the result reports that machine learning-based methods usually tend to gain higher accuracy, particularly in informal social media data.

Keywords: Sentiment Analysis, Movie Reviews, Genre-Wise Analysis, Machine Learning, LexiconBased Approach, IMDb, Twitter.