In my recent project, I focused on enhancing the analysis of Hindi legal text through Named Entity Recognition (NER). Leveraging Python, Transformers, and the Hugging Face library, I fine-tuned the MuRIL (Multilingual Representations for Indian Languages) model. The outcome was an impressive accuracy of 90.16%.
-
MuRIL Model Fine-Tuning:
- Achieved a high accuracy of 90.16% by fine-tuning the MuRIL model, specifically designed for multilingual representations in Indian languages.
-
Data Collection and Transformation:
- Collected, converted, and transformed over 100 Judicial Reports using Image-to-text conversion tools to prepare a comprehensive dataset.
-
Manual Annotation:
- Annotated more than 3000 data lines manually using various tools, including the NER Text Annotator Tool, to ensure precise and accurate training data.
-
Tools and Technologies: Python, Transformers, Hugging Face, NER Text Annotator
-
Completion Date: August 2023
This project not only showcases my proficiency in NER and language processing but also highlights my commitment to creating valuable insights in the domain of legal text analysis.
Feel free to explore the project repository for more details and insights into the NER process on legal text in Hindi. If you have any questions or feedback, don't hesitate to reach out!