The Challenge of Code-Switching: Sentiment Analysis in Multilingual Social Media Data
Sumedh Gajbhiye
This an important study as it tackles the issue of parsing huge amounts of informal, unstructured user-generated content on X and Instagram. The proposed work designs a solid pipeline for polarity classification by using supervised machine learning framework using Sentiment140 and Hinglish_Sentiment datasets. The project includes novel preprocessing and cleaning methods for social media-specific, shorthand slangs, and text that is limited by a certain number of characters. Early results show that sentiment detection can be done with a high degree of accuracy, confirmed our belief that this type of automated analysis is precise and scalable; an essential component in real-time market research, public policy tracking and helping clients navigate community mood.