Rationalism in the face of GPT hypes: Benchmarking the output of Large Language Models against human expert-curated biomedical knowledge graphs
This repository provides the data and source code for our research regarding the benchmarking the output of LLMs against human expert-curated biomedical disease-specific knowledge graphs
The different dataset and KG used in this study are located in data directory. This includes the abstracts as well as full-text documents stored in small chunks. The extracted triples using different tools are stored as an excel file for each abstract and full-text experiment, separately.
The code directory contains the scripts for extractin triples as well as querying and analysis of loaded KGs in Neo4j.