A Comparative Study of ML and NLP Models with Sentimental Analysis
Anita Pisote, Satyakam Mangate, Yash Tarde, Hiba Anis Inamdar, Shubhangi Ashok Nangare, Vishal Borate
Abstract
This paper focuses on enhancing sentiment analysis for Marathi, an underrepresented language in Natural Language Processing (NLP). While sentiment analysis has advanced significantly in languages like English and Spanish, Marathi faces challenges due to limited resources and datasets. In order to meet this, the paper adopts machine learning (ML) and NLP models specifically designed for processing Marathi text on social media platforms like YouTube, Twitter, and Instagram. One of the most important innovations of this work is the use of a correction mechanism inside the sentiment analysis pipeline. The mechanism corrects errors commonly encountered in the everyday language of social media messages. Through the use of a customized dataset with tailored lexicons and root words, the system uses more precise sentiment detection and offers accurate summaries of what is being analyzed. The study adds to the multilingual NLP discipline in terms of building a consistent sentiment analysis tool for Marathi, thereby bridging an important gap and paving the way for such advancements in other regional languages