Artificial intelligence and machine learning
Research Article
Selecting and Controlling Sense Granularity for Lexical Semantic Change Detection
Denis Vladislavovich Kokosinskii1
, Dominik Schlechtweg2, Nikolay Viktorovich Arefyev3
| 1 | Lomonosov Moscow State University, Moscow, Russia |
| 2 | University of Stuttgart, Germany |
| 3 | University of Oslo, Norway |
| 1 |
deniskokossskappa@gmail.com |
Abstract. Studying how word meanings change is a long standing problem in linguistics. We present an automatic approach that groups usages of a word into clusters corresponding to word senses and measures how the usage frequency in those senses changes between historical periods. The method follows the established procedures used to create recent human-annotated language resources (Diachronic Word Usage Graphs) and lets users adjust how coarseor fine-grained the senses should be. We also introduce a novel metric that allows to reliably evaluate the quality of the clusters, specifically tailored for the Diachronic Word Usage Graphs.
Across multiple languages, the approach performs on par with, and often better than, existing alternatives while providing clear, interpretable outputs that reveal which word senses contribute to the semantic change. (Linked article texts in English and in Russian).
Keywords: lexical semantic change detection, diachronic word usage graphs, word sense induction
MSC-2020
91F20; 68T50
For citation: Denis V. Kokosinskii, Dominik Schlechtweg, Nikolay V. Arefyev. Selecting and Controlling Sense Granularity for Lexical Semantic Change Detection. Program Systems: Theory and Applications, 2026, 17:2, pp. 103–146. (in Engl. In Russ.). https://psta.psiras.ru/2026/2_103-146.
Full text of bilingual article (PDF): https://psta.psiras.ru/read/psta2026_2_103-146.pdf (Clicking on the flag in the header switches the page language).
The article was submitted 18.02.2026; approved after reviewing 23.04.2026; accepted for publication 22.05.2026; published online 07.06.2026.
