Sriparna Saha
| Grant Category: | Fulbright-Nehru Academic & Professional Excellence Award (Research) |
| Project Title: | Unveiling the Linguistic Divide: A Unified LLM Fine-tuning for Low-Resource Languages |
| Field of Study: | Computer Science |
| Home Institution: | Indian Institute of Technology (IIT) Patna, Patna, Bihar |
| Host Institution: | University of South Carolina , Columbia, SC |
| Grant Start Month: | April 2025 |
| Duration of Grant: | Four months |
Brief Bio:
Dr. Sriparna Saha is currently serving as an Associate Professor in the Department of Computer Science and Engineering, IIT Patna, India. She has authored or co-authored more than 400 papers. Her current research interests include machine-learning, deep-learning, natural-language-processing, and biomedical-information-extraction. She is the recipient of Google-India-Women-in-Engineering-Award-2008, NASI-Young-Scientist-Platinum-Jubilee-Award-2016, BIRD Award-2016, IEI-Young-Engineers’-Award-2016, SERB-Women-in-Excellence-Award-2018, Pattern-Recognition-Letters-Editor-Award-2023, prestigious “Young-Faculty-Research-Fellowship” under Visvesvaraya-PhD-Scheme for Electronics-&-IT for-5-years (Jan 2019-Jan 2024), Humboldt-Research-Fellowship, Indo-U.S.-Fellowship-for-Women-in-STEMM-2018. She won the best paper awards in ICONIP 2023, CLINICAL-NLP workshop of COLING 2016, and Area-chair-award (Information Extraction) at IJCNLP-AACL 2023.
With her Fulbright fellowship, at University of South Carolina, Sriparna is working towards developing some unified large language models (LLM) for low-resource settings. In general, it has been shown in the recent literature that the existing LLMs are not performing well for low resource Indian languages like Bengali. This disparity raises concerns about the fairness of LLMs, as it may lead to biased outcomes and unequal access to information and resources for speakers of low-resource languages. In this project Sriparna aims to develop some LLMs for low-resource language setting by developing a scalable training approach using reinforcement learning from human feedback.