Robust Multilingual Sentiment Classification for Low-Resource Telugu and Telugu-English Code-Mixed Text

This study investigates multilingual sentiment classification for low-resource Telugu and Telugu-English code-mixed text. It focuses on classifying text into three sentiment categories: positive, negative, and neutral. The research considers challenges associated with Telugu script, Romanized Telugu, code-mixed expressions, and variations in language usage. The study explores machine learning and transformer-based approaches, including TF-IDF with Logistic Regression, IndicBERT, and XLM-RoBERTa. It examines the use of datasets such as CMTET, ACTSA, and TeSent to investigate sentiment classification across different forms of Telugu text. The objective is to evaluate the effectiveness of different approaches for sentiment analysis in low-resource multilingual settings and identify challenges related to language representation, data quality, and model performance.