Proceedings of the Annual Meeting of the Cognitive Science Society

Inflection classes partition the lexicon, classifying lexemes into groups based on shared inflectional exponence; as such, they are foundational aspects of lexical organisation. The task of identifying good partitions is non-trivial, however. We show that a modularity-based community detection method, borrowed from network theory, allows for identification of intuitive partitions at different granularities of representation. Applied to French verbs and Bosnian/Croatian/Montenegrin/Serbian nouns, the method detects inflection classes at multiple granularities and reveals (imperfect) hierarchical organisation of these. Ultimately, community detection methods facilitate more nuanced understanding of the inflectional organisation of the lexicon. Keywords: lexicon; morphology; inflection classes; hierarchical organisation; network science; community detection