Chrome Extension
WeChat Mini Program
Use on ChatGLM

From Local to Global: A Graph RAG Approach to Query-Focused Summarization

Darren Edge,Ha Trinh,Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt,Jonathan Larson

CoRR(2024)

Cited 79|Views412
No score
Abstract
The use of retrieval-augmented generation (RAG) to retrieve relevantinformation from an external knowledge source enables large language models(LLMs) to answer questions over private and/or previously unseen documentcollections. However, RAG fails on global questions directed at an entire textcorpus, such as "What are the main themes in the dataset?", since this isinherently a query-focused summarization (QFS) task, rather than an explicitretrieval task. Prior QFS methods, meanwhile, fail to scale to the quantitiesof text indexed by typical RAG systems. To combine the strengths of thesecontrasting methods, we propose a Graph RAG approach to question answering overprivate text corpora that scales with both the generality of user questions andthe quantity of source text to be indexed. Our approach uses an LLM to build agraph-based text index in two stages: first to derive an entity knowledge graphfrom the source documents, then to pregenerate community summaries for allgroups of closely-related entities. Given a question, each community summary isused to generate a partial response, before all partial responses are againsummarized in a final response to the user. For a class of global sensemakingquestions over datasets in the 1 million token range, we show that Graph RAGleads to substantial improvements over a naïve RAG baseline for both thecomprehensiveness and diversity of generated answers. An open-source,Python-based implementation of both global and local Graph RAG approaches isforthcoming at https://aka.ms/graphrag.
More
Translated text
AI Read Science
Must-Reading Tree
Example
Generate MRT to find the research sequence of this paper
Chat Paper
Summary is being generated by the instructions you defined