Global Entity Disambiguation with BERT

TL;DR

This paper introduces a global entity disambiguation model leveraging BERT, which considers both words and entities as input tokens, achieving state-of-the-art results across multiple datasets.

Contribution

It presents a novel BERT-based global ED model that sequentially resolves mentions using resolved entities as context, trained on a large Wikipedia-derived corpus.

Findings

Achieved state-of-the-art results on five ED datasets

Effectively models global context for entity disambiguation

Utilizes large-scale Wikipedia data for training

Abstract

We propose a global entity disambiguation (ED) model based on BERT. To capture global contextual information for ED, our model treats not only words but also entities as input tokens, and solves the task by sequentially resolving mentions to their referent entities and using resolved entities as inputs at each step. We train the model using a large entity-annotated corpus obtained from Wikipedia. We achieve new state-of-the-art results on five standard ED datasets: AIDA-CoNLL, MSNBC, AQUAINT, ACE2004, and WNED-WIKI. The source code and model checkpoint are available at https://github.com/studio-ousia/luke.