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.
@article{arxiv.2010.16097,
title = {Target Word Masking for Location Metonymy Resolution},
author = {Haonan Li and Maria Vasardani and Martin Tomko and Timothy Baldwin},
journal= {arXiv preprint arXiv:2010.16097},
year = {2022}
}