Target Word Masking for Location Metonymy Resolution
This addresses the problem of location metonymy resolution for natural language processing applications, offering a more efficient and generalizable method, though it is incremental as it builds on BERT.
The paper tackled location metonymy resolution by proposing an end-to-end word-level classification approach using only BERT, without external resources, and achieved state-of-the-art results on 5 datasets with large performance margins.
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.