IndexFziQ/GNN4NLP-Papers

What it is

GNN4NLP-Papers is a curated, frequently updated list of research papers that apply graph neural networks (GNNs) to natural language processing (NLP). If you work in or study language AI, this repo is a shortcut to finding the relevant conference papers and other key resources on this cross-field topic.

The two fields, briefly

• GNNs (graph neural networks): A family of machine-learning models designed to work on data shaped like a graph — nodes (e.g., words, sentences) connected by edges (e.g., dependencies, co-occurrence). GNNs learn by passing information between connected nodes.
• NLP (natural language processing): Getting computers to understand and generate human language — tasks like translation, question answering, sentiment analysis, and text classification.

This repo collects papers at the intersection: using graph structure to improve language tasks.

How it's organized

Two complementary ways, since you might search by topic or by venue:

• By topic ("Taxonomy"): 12 themed lists, each a separate file — e.g., Dialogue, Question Answering, Sentiment Analysis, Text Classification, Text Generation, Knowledge Graph, Information Extraction, plus Theory and Visual Question Answering. A notable recent addition is "LLM and Graph", added in the Jan. 2026 update to reflect work on large language models and graph retrieval-augmented generation (GraphRAG).
• By conference ("According to Conference"): 10 files listing papers grouped by the venue they were published at — ACL, EMNLP, NAACL, COLING, ICLR, ICML, KDD, WWW, CIKM, ICCV. The README states the repo overall covers these plus IJCAI, AAAI, NIPS, and SIGIR.

What's inside (beyond the lists)

• Update log: The README tracks each refresh — e.g., May 2022 added ACL-2022 papers, Jan. 2026 added LLM and GraphRAG content.
• Links to comprehensive GNN paper lists: Three well-known external repositories (e.g., thunlp/GNNPapers) for readers who want a broader graph-focused literature review.
• Tutorials: Two pointers — a 2019 EMNLP tutorial on GNNs-for-NLP, and Stanford's CS224W course on graph machine learning.
• Tools: Links to two popular GNN libraries — Deep Graph Library (DGL) and PyTorch Geometric (PyG).
• PhD theses: Four doctoral theses related to graph-structured representations in NLP or deep learning (e.g., by authors from Alberta, Amsterdam, IISc, and Cambridge).

Bottom line

It's a roadmap, not a tool — no code to install, no models to run. Its value is as a reliable, actively maintained map of where to find relevant papers and learning material if you want to apply graph methods to language tasks, or understand the state of the field. The recent 2026 update signals the maintainer is actively tracking where GNNs meet modern LLMs and retrieval-augmented generation.