A Transfer Learning Approach to the Analysis of Sentiment for the Kannada Language Based on Indic-BERT and XLM-RoBERTa-Base
Yathish Poojary, Dhanusha, B. Ashwath Rao, Musica Supriya, VG Narendra · 2026
Sentiment analysis or opinion mining is one of the major tasks of NLP (Natural Language Processing). In this paper, the different aspects of design, implementation, and evaluation of an artificial sentiment analysis system for movie and product reviews are considered. A contrary approach is used that does fine-grain analysis to decide the orientation of the reviewer&s;s feelings and the strength of feelings concerning different components of a film and product with most of the work concentrating only on deciding sentiments like positive, negative, and neutral. Sentiment analysis is that part of the field that understands and extracts opinions from reviews. The analysis procedure involves text analytics, natural language processing, computational linguistics, and polarity classification. We developed the supervised machine learning model for analyzing sentiments using the XLM-RoBERTa-base model. This paper studies how to apply and describes the performances of different classification algorithms for classes related to film and product reviews, and achieve a 71.9% accuracy and RMSE of 1.069. Whereas, with 70% precision, 70% of all cases predicted as positive by the model turned out to be actually positive. Furthermore, 68% of all real positive cases were accurately identified, according to the recall value of 68%. These outcomes demonstrate the model&s;s efficacy in sentiment analysis tasks and offer insightful information for natural language processing applications.