Leveraging Inter-Chunk Interactions for Enhanced Retrieval in Large Language Model-Based Question Answering

TL;DR

This paper introduces IIER, a novel retrieval framework that leverages inter-chunk interactions and a graph-based approach to improve external knowledge retrieval for large language model question answering, especially in multi-document scenarios.

Contribution

The paper proposes a new retrieval framework that models internal document chunk interactions and constructs a unified graph to enhance retrieval accuracy and reasoning in LLM-based QA.

Findings

IIER outperforms baseline methods across four datasets.

Modeling inter-chunk interactions improves retrieval relevance.

Graph-based evidence chaining enhances reasoning capabilities.

Abstract

Retrieving external knowledge and prompting large language models with relevant information is an effective paradigm to enhance the performance of question-answering tasks. Previous research typically handles paragraphs from external documents in isolation, resulting in a lack of context and ambiguous references, particularly in multi-document and complex tasks. To overcome these challenges, we propose a new retrieval framework IIER, that leverages Inter-chunk Interactions to Enhance Retrieval. This framework captures the internal connections between document chunks by considering three types of interactions: structural, keyword, and semantic. We then construct a unified Chunk-Interaction Graph to represent all external documents comprehensively. Additionally, we design a graph-based evidence chain retriever that utilizes previous paths and chunk interactions to guide the retrieval process. It identifies multiple seed nodes based on the target question and iteratively searches for relevant chunks to gather supporting evidence. This retrieval process refines the context and reasoning chain, aiding the large language model in reasoning and answer generation. Extensive experiments demonstrate that IIER outperforms strong baselines across four datasets, highlighting its effectiveness in improving retrieval and reasoning capabilities.