Researchers use NLP, knowledge graphs, and topic modeling to extract and classify medical knowledge from translated ancient Indian texts

A new arXiv preprint describes a computational pipeline that applies Natural Language Processing (NLP) techniques to translated ancient Indian medical texts, including the Sushruta Samhita, to extract and organize domain knowledge. The authors use Named Entity Recognition (NER) to identify structured entities such as diseases, treatments, researchers, and medicinal plants.

To categorize underlying medical themes, the team applies BERTopic modeling, which clusters passages into thematic groups based on content. They then construct a knowledge graph in Neo4j to semantically represent relationships among the extracted entities, enabling network-based analysis and knowledge retrieval.

The authors report that their approach facilitates computational organization of Ayurvedic medical wisdom and aims to close the gap between traditional textual formats and contemporary data-driven research methods. They note applications in historical text analysis, medical informatics, and digital humanities.

The research was presented at the National Conference on 'Reimagining LIS Education: Integrating Indian Knowledge Systems with NEP 2020' in March 2025, organized by the Tata Institute of Social Sciences and the Indian Association of Teachers of Library and Information Science, where it received the Best Paper Award.