Explaining and Improving BERT Performance on Lexical Semantic Change Detection
This addresses a specific bottleneck in NLP for researchers in lexical semantics, offering an incremental improvement.
The paper tackled the problem of why token-based models like BERT underperform in lexical semantic change detection, finding that orthographic information in BERT vectors reduces performance, and by mitigating this, they improved BERT's performance by 15% on SemEval-2020 Task 1.
Type- and token-based embedding architectures are still competing in lexical semantic change detection. The recent success of type-based models in SemEval-2020 Task 1 has raised the question why the success of token-based models on a variety of other NLP tasks does not translate to our field. We investigate the influence of a range of variables on clusterings of BERT vectors and show that its low performance is largely due to orthographic information on the target word, which is encoded even in the higher layers of BERT representations. By reducing the influence of orthography we considerably improve BERT's performance.