Harnessing Large Language Models for Graph Learning

Traditionally, graph neural networks (GNNs) have been the workhorse for graph learning tasks, achieving impressive results. However, recent research explores the exciting potential of large language models (LLMs) in this domain.

In this chapter, we’ll delve into the intersection of LLMs and graph learning, exploring how these powerful language models can enhance graph-based tasks. We’ll begin with an overview of LLMs, followed by a discussion of the limitations of GNNs and the motivations for incorporating LLMs. Then, we’ll explore various approaches for utilizing LLMs in graph learning, the intersection of retrieval-augmented generation (RAG) with graphs, and explain the advantages and challenges associated with this integration.

In this chapter, we’ll explore the following topics:

• Understanding LLMs
• Textual data in graphs
• LLMs for graph learning
• Integrating RAG with graph learning...