Does BERT Make Any Sense? Interpretable Word Sense Disambiguation with Contextualized Embeddings

23 Sep 2019  ·  Gregor Wiedemann, Steffen Remus, Avi Chawla, Chris Biemann ·

Contextualized word embeddings (CWE) such as provided by ELMo (Peters et al., 2018), Flair NLP (Akbik et al., 2018), or BERT (Devlin et al., 2019) are a major recent innovation in NLP. CWEs provide semantic vector representations of words depending on their respective context. Their advantage over static word embeddings has been shown for a number of tasks, such as text classification, sequence tagging, or machine translation. Since vectors of the same word type can vary depending on the respective context, they implicitly provide a model for word sense disambiguation (WSD). We introduce a simple but effective approach to WSD using a nearest neighbor classification on CWEs. We compare the performance of different CWE models for the task and can report improvements above the current state of the art for two standard WSD benchmark datasets. We further show that the pre-trained BERT model is able to place polysemic words into distinct 'sense' regions of the embedding space, while ELMo and Flair NLP do not seem to possess this ability.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Word Sense Disambiguation SemEval 2007 Task 17 kNN-BERT F1 60.94 # 8
Word Sense Disambiguation SemEval 2007 Task 17 kNN-BERT + POS (training corpus: SemCor) F1 63.17 # 7
Word Sense Disambiguation SemEval 2007 Task 7 kNN-BERT F1 81.20 # 9
Word Sense Disambiguation SemEval 2007 Task 7 kNN-BERT + POS (training corpus: WNGT) F1 85.32 # 3
Word Sense Disambiguation SensEval 2 Lexical Sample kNN-BERT F1 76.52 # 1
Word Sense Disambiguation SensEval 3 Lexical Sample kNN-BERT F1 80.12 # 1

Methods