Sentiment Analysis in Code-Mixed Video Comments Using Machine Learning on a Hinglish Dataset

Given the fluency of user-created information across languages, colloquialisms, and cultural nuances the task of sentiment analysis in code-mixed video comments is challenging. In this study, sentiment analysis of bilingual (English-Hindi) comments from YouTube, Instagram, and other video-sharing platforms has received considerable importance. To address these problems, we created a large set of code-mixed comments collected from various sites. Employing the voting technique, we also combined machine learning algorithms, including the Random Forest and Logistic Regression algorithms, along with NLP to form an ensemble machine learning model. This made it possible to extract subtleties from the bilingual text with a superb $\mathbf{9 1 \%}$ sentiment classification accuracy. In this work, it was revealed that the use of advanced models for code-mixed text, audience engagement tactics recommended content tactics, and content moderation strategies in well-used video platforms may be significantly improved. The findings show that audience engagement tactics, targeted recommendations, and content moderation on wellknown video platforms may all be considerably enhanced by utilizing sophisticated models designed for code-mixed text. This work lays the groundwork for future multilingual sentiment analysis research, and our dataset can be an invaluable tool for developing the discipline.