← all repositories
roomylee/awesome-relation-extraction

A reading list for teaching machines who knows whom

A curated index of papers, code, and datasets for relation extraction, the NLP task of spotting connections between named entities.

awesome-relation-extraction
Velocity · 7d
+0.4
★ / day
Trend
steady
star history

What it does This repository is a bibliography with links — nothing more, nothing less. It catalogs research on relation extraction, the subfield of NLP concerned with identifying how entities in text relate to each other (think “founded_by” or “located_in”). The list spans surveys, conference papers from ACL and EMNLP, code repositories, datasets, and a handful of video lectures.

The interesting bit The curation is methodical rather than merely chronological. Papers are grouped by technique — CNN-based, RNN-based, graph neural networks, distant supervision, language models — which makes it useful for tracing how the field evolved from feature engineering to pre-trained transformers. The maintainer also links to their own paper reviews and reproduction code where available.

Key highlights

  • Coverage from 2012 (Socher’s matrix-vector recursive model) through 2021 (GNN and knowledge-graph hybrids)
  • Distant supervision section includes the foundational Mintz et al. 2009 paper and its neural successors
  • Links to active frameworks and systems, not just archived research
  • Includes biomedical-specific resources (CoNLL 2017 papers) alongside general-domain work
  • Maintainer accepts pull requests and has contributed personal code reviews to several entries

Caveats

  • Some paper links point to author-hosted PDFs or preprints that may drift or break
  • A typo in one entry (“parer” instead of “paper” for RECON) suggests light proofreading
  • The list appears to trail off mid-entry at “RESIDE:” — whether truncated in the source or abandoned is unclear

Verdict Worth bookmarking if you’re entering relation extraction research or need to map the technique landscape from 2012–2021. Skip it if you want executable tools or up-to-the-minute LLM-era coverage; this is a historical index, not a living benchmark.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.