Reproduce “Collaborative Knowledge Graph Fusion by Exploiting the Open Corpus” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yue Wang, Yao Wan, Lu Bai, Lixin Cui, Zhuo Xu, Ming Li, Philip S. Yu, Edwin R Hancock
A collaborative framework is introduced to enrich knowledge graphs by alternately leveraging Joint Event Extraction (JEE) from open text and Knowledge Graph Fusion (KGF), fostering mutual improvement between knowledge extraction and graph quality. The system, featuring a Benchmark-based Supervision JEE and a Relation Alignment-based OKGF, improved JEE F1-scores on ACE 2005 (e.g., 88.6% for entity extraction) and achieved 100% Hit@10 for triple prediction on FB15K.
Key insights
- The alternating, collaborative interaction between knowledge extraction (JEE) and knowledge graph fusion (KGF) tasks enables continuous mutual enhancement, where improved graph quality guides better extraction, and better extraction enriches the graph.
- The Translated Relation Alignment Score (TRAS) effectively aligns diverse relation mentions from open text with a prior KG's schema by combining textual and translated semantic similarities, overcoming a key challenge in integrating heterogeneous information.
- Benchmark-based supervision provides structural context from the evolving KG to guide the text extraction process, reducing errors and improving the relevance of extracted knowledge.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.