Using Synchronic Definitions and Semantic Relations to Classify Semantic Change Types

TL;DR

This work tackles the problem of not only detecting when words change meaning but also identifying the type of change they undergo. It introduces a definition-based, WordNet-driven approach that maps diachronic change types to synchronic relations (generalization, specialization, co-hyponymy, antonymy, homonymy) and trains a classifier on WordNet definitions. The authors release the LSC-CTD benchmark, digitized from Blank's taxonomy with concise sense definitions, and show that incorporating change-type information improves both graded WiC and binary LSC tasks. They demonstrate that homonymy is the most distant relation in human judgments and achieve state-of-the-art or competitive results on SemEval-2020 Task 1 when combining their definitions-based signal with existing models.

Abstract

There is abundant evidence of the fact that the way words change their meaning can be classified in different types of change, highlighting the relationship between the old and new meanings (among which generalization, specialization and co-hyponymy transfer). In this paper, we present a way of detecting these types of change by constructing a model that leverages information both from synchronic lexical relations and definitions of word meanings. Specifically, we use synset definitions and hierarchy information from WordNet and test it on a digitized version of Blank's (1997) dataset of semantic change types. Finally, we show how the sense relationships can improve models for both approximation of human judgments of semantic relatedness as well as binary Lexical Semantic Change Detection.

Figures (7)

• Figure 1: Confusion matrix for predictions of sense relationships on the WordNet testset.
• Figure 2: Confusion matrix for predictions of sense relationships on the LSC-CTD Benchmark.
• Figure 3: Ratio of human judgments for each class.
• Figure 4: SGD TF-IDF.
• Figure 5: roberta-base.