SlaViCoN: Slavic Verbs in Computational Networks

SlaViCoN uses computational network analysis (CNA) to investigate the structure of Slavic verbal morphology, focusing on networks of prefixes, roots, and theme vowels. Drawing on methods from network science, the project evaluates competing theories of morphology and addresses fundamental questions about the nature of morphemes and their relationship to syntax and semantics: Are morphemes pairings of form and meaning, autonomous morphological units, or realizations of grammatical features? It also examines the productivity of verbal morphemes, the distinction between lexical and superlexical prefixes, and variation across Slavic verbal systems.

The project draws primarily on the WeSoSlaV database (Arsenijević et al., 2024), which contains 8,300 of the most frequent verbs in Bosnian/Croatian/Montenegrin/Serbian (BCMS) and Slovenian, annotated for more than 40 properties, including affixation, prosody, and aspect. It constructs networks in which morphemes are connected according to shared distributional properties – for example, prefixes are linked when they combine with the same verbs – and analyzes them using measures such as network density, clustering, centrality, and cosine similarity. The predictions of competing morphological theories are evaluated by comparing networks constructed from different combinations of syntactic, semantic, and phonological features. Cross-linguistic comparisons examine whether corresponding networks in BCMS and Slovenian differ significantly in their topology.

SlaViCoN is the first project to apply CNA to a manually annotated morphological dataset of this complexity and richness. By incorporating a broad range of theoretically relevant properties into network models, it considerably expands the potential of CNA to address foundational questions in morphological theory. At the same time, the project tests the method on exceptionally rich linguistic data, opening new avenues for methodological development.