Fine-Tuned BERT Based Multilingual Model for Named Entity Recognition in Native Indian Languages
G Bharathi Mohan, R Prasanna Kumar, R Elakkiya, Mukhtesh Venkata Sri Sai Pendem
Abstract
Accurate recognition of named entities is essential in real-world applications such as information extraction, document classification, sentiment analysis, and machine translation.The fine-tuned BERT model demonstrates exceptional performance in multilingual NER tasks by leveraging language-specific opti-mization, cross-lingual transfer, and context understanding. In contrast, other models applied to multilingual data may exhibit limitations. The research utilizes the WikiAnn dataset, including languages like Tamil, Malayalam, Bengali, and Marathi, to train the fine-tuned BERT model.The system achieved a precision of 95.3% and a recall of 94.8% for the Bengali language, outperforming the generic model used for Indian languages.It demonstrates superior accuracy and effectiveness in deciphering Bengali text.This research provides a robust solution for accurate named entity recognition across multiple languages, benefiting real-world applications.