Back to feed
arXiv cs.CL
arXiv cs.CL
7/23/2026
Lightweight Person-Place Relation Extraction from Historical Newspapers with Dependency Graphs and Proximity Features

Lightweight Person-Place Relation Extraction from Historical Newspapers with Dependency Graphs and Proximity Features

Short summary

A team at DS@GT tackled the HIPE-2026 shared task for person-place relation extraction from multilingual historical newspapers using lightweight dependency-graph features and small ML ensembles—no pretrained LMs at the classification stage. Their best run ranked 3rd on efficiency with macro recall of 0.5142. Key findings: character distance alone captures most signal, and document-grouped cross-validation is essential to avoid 25-37 point score inflation from entity recurrence across documents.

  • •Lightweight system (<847K params) for person-place relation extraction ranked 3rd on efficiency in HIPE-2026
  • •Minimum character distance captures most classification signal; additional features yield inconsistent gains
  • •Document-grouped cross-validation is critical—pair-level splits inflate scores by 25-37 percentage points

Generated with AI, which can make mistakes.

#research-breakthrough
Read full article at arXiv cs.CL

Is this a good recommendation for you?

Comments

Failed to load comments. Please try again.

Explore more

Profile-Graph Memory for LLM Agents: Implicit Cross-Entity Traversal through Narrative Profiles
arXiv CS.AI
LLM-INSTRUCT at UZH Shared Task 2026: Constraint-Aware Retrieval and Selective Debate for Paragraph-Level Argument Mining
arXiv cs.CL
Democratizing AI with Small Language Models: Structured Benchmarking and Parameter-Efficient Fine-Tuning for Local Deployment
arXiv CS.AI
HG-RAG: Hierarchy-Guided Retrieval-Augmented Generation for Structured Knowledge Graphs
arXiv CS.AI