Prajat Paul
Doctoral Student | 3rd Year
“Doctoral student researching speech-based clinical NLP and EHR systems for low-resource languages”
Biography
Prajat Paul is a doctoral student at the Graduate School of Information Science and Electrical Engineering, Kyushu University, Japan, supported by the Monbukagakusho (MEXT) Scholarship. He completed his B.Tech in Electronics and Communication Engineering from NIT Durgapur, India. His research sits at the intersection of automatic speech recognition (ASR), clinical natural language processing, and electronic health records (EHR), with a focus on low-resource languages. He is also co-founder of KenkoMap/SHGC, a health-tech startup within the PHC ecosystem, funded by JST PARKS.
Research Summary
This research explores a speech-based healthcare information extraction system as a supplementary EHR module for low-resource languages (LRLs), specifically Bangla. It focuses on extracting and classifying key medical entities from doctor-patient conversations to assist people with text illiteracy, improve EHR comprehensiveness, and support clinical decision-making through a pre-consultation AI tool. Key objectives: 1. Assess existing ASR tools and factors affecting LRL performance 2. Extract and classify clinical information from doctor-patient conversations for EHR storage 3. Develop a pre-consultation AI to retrieve chief complaints with minimal human intervention
Legacy Archive Profile
Imported record from the previous Social Tech Lab website.
Mr. Prajat Paul is currently a Doctoral Student at the Graduate School of Information Science and Electrical Engineering at Kyushu University, Japan. He pursued his undergraduate education at the National Institute of Technology Durgapur (India), majoring in Electronics and Communication Engineering under the ICCR scholarship provided by the Government of India. His current research endeavors at the SocialTech Lab are being supported by the Monbukagakusho (MEXT) Scholarship funded by the Government of Japan.
Mr. Paul’s research aims to explore the implementation of a speech-based healthcare information extraction system as a supplementary module of the conventional Electronic Health Record (EHR) system. Focusing on low-resource languages (LRLs), specifically Bangla, this system aims to assist people with text illiteracy by extracting and classifying important medical information from spoken language. The goal is to create more comprehensive EHRs, leading to improved healthcare efficiency and offering doctors predictive support for clinical decision-making.
Research Objective: 1. Observe the current status and efficiency of existing ASR tools and the reasons that affect their performance for Low-Resource Languages (LRL) 2. Extract information from the Doctor-Patient Conversation and assess its usability in storing 3. Ideate a medical pre-consultation AI tool that can retrieve Chief Complaint (CC) and initiate the process of receiving healthcare with minimal human intervention requirements
Publications
Towards Inclusive Digital Health: An Architecture to Extract Health Information from Patients with Low-Resource Language
Paul, P., Bouh, M. M., & Ahmed, A. (2024, January). Towards Inclusive Digital Health: An Architecture to Extract Health Information from Patients with Low-Resource Language. In BIOSTEC (2) (pp. 754-760).
A Comprehensive Study on Bangla Automatic Speech Recognition Systems
Paul, P., Bouh, M. M., Hossain, F., & Ahmed, A. (2023, November). A Comprehensive Study on Bangla Automatic Speech Recognition Systems. In 2023 2nd International Conference on Frontiers of Communications, Information System and Data Science (CISDS) (pp. 72-77). IEEE.
Investigating Sibilant Fricative Representation in Bangla Telemedicine Speech: A Cost-Aware Sampling Rate Optimization Study
Paul, P., Bouh, M. M., Shah, M. V., Hossain, F., & Ahmed, A. (2026). Investigating Sibilant Fricative Representation in Bangla Telemedicine Speech: A Cost-Aware Sampling Rate Optimization Study. Signals, 7(3), 44. https://doi.org/10.3390/signals7030044
Achievements
JST PARKS Startup Ecosystem Co-creation Program for the New Industry Creation Fund for University Startups
This is a startup fund received in October 2024 from JST for a student project exploring the initiation of commercialization into becoming a startup. The project, currently known as KenKoMap, is a health data visualization and management tool that connects with existing Electronic Health Record (EHR) systems and visualizes data categorized based on who is viewing the data. Total amount of pre-seed funds received: 1M Yen (Step 1, Oct 2024-Sept 2025), 3M Yen (Step 2, Oct 2025-Sept 2026)