ReTAG: Retrieval-Enhanced, Topic-Augmented Graph-Based Global Sensemaking

2025-09-30

AI Summary

Global meaning construction—i.e., cross-corpus holistic information integration and reasoning—faces challenges including retrieval incompleteness, topic over-generalization, and high inference overhead. To address these, we propose ReTAG: a framework that first constructs topic-enhanced subgraphs via topic modeling, then integrates retrieval mechanisms for precise key-information localization, and finally leverages synergistic graph neural networks and generative models to perform multi-hop reasoning and answer synthesis. Compared to existing graph-based approaches, ReTAG achieves state-of-the-art performance on benchmarks such as HotpotQA, significantly improving multi-hop question answering accuracy while reducing inference latency by 42%–68%. Its core contribution lies in the first deep coupling of topic-aware subgraph construction with retrieval augmentation—effectively balancing semantic focus and computational efficiency—and establishing a scalable new paradigm for global reasoning over large-scale corpora.

Technology Category

• Knowledge Representation and Reasoning: Computational Complexity of Reasoning
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Recent advances in question answering have led to substantial progress in tasks such as multi-hop reasoning. However, global sensemaking-answering questions by synthesizing information from an entire corpus remains a significant challenge. A prior graph-based approach to global sensemaking lacks retrieval mechanisms, topic specificity, and incurs high inference costs. To address these limitations, we propose ReTAG, a Retrieval-Enhanced, Topic-Augmented Graph framework that constructs topic-specific subgraphs and retrieves the relevant summaries for response generation. Experiments show that ReTAG improves response quality while significantly reducing inference time compared to the baseline. Our code is available at https://github.com/bykimby/retag.

Problem

• Addresses global sensemaking challenges in corpus-wide question answering
• Enhances graph-based methods with retrieval mechanisms and topic specificity
• Reduces high inference costs while improving response quality

Innovation

• Retrieval-enhanced graph framework for global sensemaking
• Constructs topic-specific subgraphs for focused reasoning
• Retrieves relevant summaries to generate accurate responses