Target Word Masking for Location Metonymy Resolution

TL;DR

This paper introduces a BERT-based end-to-end method for location metonymy resolution that outperforms existing models and does not rely on external resources or handcrafted features.

Contribution

The authors present a novel word-level classification approach using only BERT, eliminating the need for external lexical resources or taggers.

Findings

Achieves state-of-the-art results on 5 datasets

Outperforms conventional BERT models and benchmarks

Generalizes well to unseen data

Abstract

Existing metonymy resolution approaches rely on features extracted from external resources like dictionaries and hand-crafted lexical resources. In this paper, we propose an end-to-end word-level classification approach based only on BERT, without dependencies on taggers, parsers, curated dictionaries of place names, or other external resources. We show that our approach achieves the state-of-the-art on 5 datasets, surpassing conventional BERT models and benchmarks by a large margin. We also show that our approach generalises well to unseen data.