Reproduce “Talking to GDELT Through Knowledge Graphs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Audun Myers, Max Vargas, Sinan G. Aksoy, Cliff Joslyn, Benjamin Wilson, Lee Burke, Tom Grimes
This research evaluates Retrieval Augmented Generation (RAG) approaches by integrating Large Language Models with both text and knowledge graph (KG) representations derived from the GDELT dataset. Researchers constructed an ontology-based KG from GDELT, and two LLM-generated KGs from news articles, demonstrating that the ontology-based KG provided superior answers for aggregate queries, while traditional RAG and GraphRAG excelled at fine-grained, document-specific details.
Key insights
- Ontology-based KGs, meticulously derived from structured data, offer a robust solution for answering complex, aggregate, and quantitative questions, surpassing text-based RAG.
- Current LLM-generated KGs from unstructured text exhibit significant structural inconsistencies, poor ontological adherence, and entity resolution issues, limiting their utility for precise question-answering.
- Automated subgraph retrieval techniques, such as G-Retriever, require further development to reliably extract comprehensive and accurate information from complex KGs.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.