This repository continues our previous work titled "Towards Building an RDF-based Deep Document Model and Retrieval-Augmented Generation System for Enhanced Question Answering with Large Language Models".
The primary focus of this repository is:
- Integration of the PARSE+DDM pipelines to build a automated framework for Knowledge graph construction and information retrieval.
- Conducting experiments with StructRAG, a Structure-Aware Retrieval-Augmented Generation Framework, which leverages a Scholarly Knowledge Graph
To enable:
- Structurally-aware question answering by incorporating hierarchical document structures.
- Diverse and path-enhanced responses through enhanced retrieval mechanisms.
- Accurate and high-quality answers by leveraging structural metadata and scholarly knowledge graphs.