Rajvee Sheth

Natural Language Processing

Developing robust computational frameworks for multilingual text understanding, with emphasis on advanced annotation methodologies and scalable evaluation systems that enable interpretable and secure NLP solutions across diverse linguistic contexts.

Code-Mixing

Redefining code-mixed NLP with curated Hindi-English datasets, robust annotation frameworks and comprehensive evaluation on existing state-of-the-art LLMs to improve language understanding.

Curating and constructing benchmarks and development of ML models for low level NLP tasks in Hindi-English code-mixing

Funded by ANRF - This comprehensive project encompasses the development of benchmarks and Annotation Framework for Hindi-English code-mixing research. Key deliverables include:

COMI-LINGUA Dataset: A large-scale, expert-annotated dataset for multitask NLP in Hindi-English code-mixing, covering tasks like Matrix Language Identification, POS Tagging, and Named Entity Recognition. Available at: Hugging Face.

COMMENTATOR Portal: A code-mixed multilingual text annotation framework designed to facilitate the annotation and analysis of Hindi-English texts. This tool supports the creation and management of annotated datasets for research purposes. Available at: GitHub Repo.

More about the work: Project Website