Unsupervised sentiment analysis of Hindi reviews using MCDM and game model optimization techniques

In this study, we develop a novel multi-criteria decision-making (MCDM) and game-theoretic mathematical framework for analyzing review sentiment. The ratings and feedback of Hindi-speaking reviewers have been collected in an interactive database. This was accomplished by initially employing a game-theoretic approach to evaluating each review, based on the multi-objective optimization technique Complex Proportional Assessment (COPRAS), and then letting the two participants play the game until they reached a Nash equilibrium. Next, we extract the sentiment label from the inferred Hindi review dataset. To gauge the overall sentiment of a review, we classify it as either good, negative, or neutral. We analyze reviews by assigning a star rating and polarity score to comments written in the HindiSentiWordNet (HSWN) lexicon. To classify unstructured sentiment, we offer a model that optimizes for both polarity and rating scores. Our proposed model achieves comparable results to state-of-the-art models, as evidenced by experimental results on three widely used Hindi review datasets. We also use statistical analysis to determine the importance of the findings. The proposed MCDM Model ensures sound reasoning and consistency. In simulations, the proposed algorithm is seen to outperform the baseline and state-of-the-art approaches. This new benchmark for sentiment analysis was achieved by incorporating the rating and polarity score of the Hindi reviews into the MCDM and game model. Additionally, our technique is very generalizable and can do sentiment analysis across many different domains.