Nested Named Entity Recognition as Building Local Hypergraphs

Named entity recognition is a fundamental task in natural language processing. Based on the sequence labeling paradigm for flat named recognition, multiple methods have been developed to handle nested structures. However, they either require fixed order or introduce complex hypergraphs. To tackle this problem, we propose novel model Local Hypergraph Builder Network (LHBN) that builds simpler local hypergraphs capture entities instead of single full-size hypergraph. The proposed has three main properties: (1) share boundaries are captured same (2) boundary information enhanced by building (3) can be built bidirectionally take advantage identification direction preference different entities. Experiments illustrate our outperforms previous state-of-the-art four widely used datasets: ACE04, ACE05, GENIA, and KBP17. code available at https://github.com/yanyk13/local-hypergraph-building-network.git.