Graph Convolutional Networks for Named Entity Recognition

TL;DR

This paper explores the use of Graph Convolutional Networks (GCNs) with dependency trees to improve Named Entity Recognition (NER), showing that sentence grammar enhances performance without complex feature engineering.

Contribution

It introduces GCNs for NER and demonstrates their effectiveness across different architectures, highlighting the importance of sentence grammar.

Findings

GCNs improve NER performance on OntoNotes dataset

Sentence grammar positively influences NER accuracy

No heavy feature engineering needed

Abstract

In this paper we investigate the role of the dependency tree in a named entity recognizer upon using a set of GCN. We perform a comparison among different NER architectures and show that the grammar of a sentence positively influences the results. Experiments on the ontonotes dataset demonstrate consistent performance improvements, without requiring heavy feature engineering nor additional language-specific knowledge.