Exploring the effect of semantic diversity on boundary permeability in verb/noun heterosemy using deep contextualized word embedding

Abstract This study investigates the relationship between semantic diversity and boundary permeability in verb/noun heterosemous words in English. Using sense data from the Oxford Dictionaries API and the Corpus of Historical American English enhanced by contextualized word embedding, we quantified semantic diversity using Shannon entropy and measured boundary permeability as the relative proportion of verb/noun uses. Results reveal a significant positive effect of overall semantic diversity on boundary permeability, i.e., words that have more senses or balanced sense distributions show greater cross-categorical flexibility. Furthermore, we find asymmetric effects between noun and verb semantic diversity: noun-sense diversity positively affects permeability, while no meaningful effect was detected for verb-sense diversity. This asymmetry, combined with our finding that higher permeability is driven by increased verb usage, provides strong quantitative evidence for a directional preference in English conversion from noun to verb. To our knowledge, this study offers the first large-scale empirical support for theoretical accounts of directionality in English conversion, demonstrating that semantic diversity functions as a critical indicator of a word’s conceptual flexibility.