HyperG: Hypergraph-Enhanced LLMs for Structured Knowledge

2025-02-25

AI Summary

Large language models (LLMs) struggle to model complex topological relationships and handle sparsity in structured knowledge (e.g., HTML web pages). To address this, we propose Prompt-aware Hypergraph Learning (PHL), a novel framework integrating LLM-driven generative data augmentation with hypergraph neural networks. PHL jointly encodes semantic and structural relationships via a prompt attention mechanism and introduces structured embeddings alongside end-to-end joint optimization. Crucially, PHL is the first approach to synergistically combine prompt engineering and hypergraph learning for structured knowledge modeling. Evaluated on two downstream tasks—structured document understanding and sparse web element classification—PHL achieves significant improvements over state-of-the-art methods. Notably, it demonstrates superior robustness and generalization under low-resource and highly sparse conditions, validating its effectiveness in capturing both fine-grained semantics and higher-order structural dependencies inherent in hierarchical, semi-structured data.

Technology Category

• Machine Learning: Structured Learning
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion
• Natural Language Processing: Prompt Engineering / Prompting

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

Given that substantial amounts of domain-specific knowledge are stored in structured formats, such as web data organized through HTML, Large Language Models (LLMs) are expected to fully comprehend this structured information to broaden their applications in various real-world downstream tasks. Current approaches for applying LLMs to structured data fall into two main categories: serialization-based and operation-based methods. Both approaches, whether relying on serialization or using SQL-like operations as an intermediary, encounter difficulties in fully capturing structural relationships and effectively handling sparse data. To address these unique characteristics of structured data, we propose HyperG, a hypergraph-based generation framework aimed at enhancing LLMs' ability to process structured knowledge. Specifically, HyperG first augment sparse data with contextual information, leveraging the generative power of LLMs, and incorporate a prompt-attentive hypergraph learning (PHL) network to encode both the augmented information and the intricate structural relationships within the data. To validate the effectiveness and generalization of HyperG, we conduct extensive experiments across two different downstream tasks requiring structured knowledge.

Problem

• Enhance LLMs for structured knowledge
• Address sparse data and structural relationships
• Validate HyperG in downstream tasks

Innovation

• Hypergraph-based generation framework
• Augments sparse data contextually
• Prompt-attentive hypergraph learning network