Exploring Category Structure with Contextual Language Models and Lexical Semantic Networks

TL;DR

This paper improves the prediction of category typicality scores by using better probing methods for contextual language models like BERT, emphasizing disambiguation and combining lexical semantic networks for enhanced results.

Contribution

It introduces improved probing techniques for BERT, highlights the role of polysemy and disambiguation, and demonstrates the complementary benefits of combining lexical semantic networks with language models.

Findings

BERT-based methods outperform previous models in typicality prediction.

Disambiguation significantly improves model performance.

Combining WordNet with BERT yields better predictions than either alone.

Abstract

Recent work on predicting category structure with distributional models, using either static word embeddings (Heyman and Heyman, 2019) or contextualized language models (CLMs) (Misra et al., 2021), report low correlations with human ratings, thus calling into question their plausibility as models of human semantic memory. In this work, we revisit this question testing a wider array of methods for probing CLMs for predicting typicality scores. Our experiments, using BERT (Devlin et al., 2018), show the importance of using the right type of CLM probes, as our best BERT-based typicality prediction methods substantially improve over previous works. Second, our results highlight the importance of polysemy in this task: our best results are obtained when using a disambiguation mechanism. Finally, additional experiments reveal that Information Contentbased WordNet (Miller, 1995), also endowed with disambiguation, match the performance of the best BERT-based method, and in fact capture complementary information, which can be combined with BERT to achieve enhanced typicality predictions.