Multi-domain aspect-oriented sentiment analysis for movie recommendations using feature extraction

Abstract

Sentiment analysis is a well-recognized research field that has acknowledged significant attention in recent years. Researchers have made extensive efforts in employing various methodologies to explore these domains. Sentiment classification plays a fundamental role in natural language processing (NLP). However, studies have shown that sentiment classification models heavily depend on the specific domain. In the context of movie industry, where the demand for reliable movie reviews is high and not all movies are of exceptional quality and worthy of viewers time. Therefore, people depend on movie reviews before watching a movie. This explores the use of data from various domains to improve classification performance within each domain, addressing the difficulty of multi-domain sentiment classification in natural language processing. Therefore, it is crucial to effectively utilize shared sentiment knowledge across different domains for real-world applications. To solve these issues, a multi-domain aspect-oriented sentiment analysis for movie recommendation using feature extraction techniques. The main contribution of this work is to eliminate the time for users to go through a lengthy list of movies to make their decision. The novelty of this work is analysis of different movie genres, TV shows genres with accurate results. The presented approach's performance is validated by evaluating various metrics, including precision, recall, mean square error (MSE) and F1-score.