Distributional Traces of Paucality in Ukrainian: A Contextual Embedding Analysis of Noun–Numeral Constructions
Distributional Traces of Paucality in Ukrainian: A Contextual Embedding Analysis of Noun–Numeral Constructions

Summary/Abstract: This study investigates the distributional organization of Ukrainian noun– numeral constructions using contextual word embeddings. Focusing on cardinal numerals, ordinal numerals, and bare nominal forms, we examine whether historically grounded distinctions in the Indo-European number system are reflected in contemporary usage patterns. Contextual embeddings of Ukrainian nouns were extracted from largescale corpus data and analyzed using dimensionality reduction and cosine similarity measures. Across multiple projections, numeral type emerges as the dominant organizing dimension of the embedding space, with a robust geometric separation between cardinal and ordinal numerals. Within the cardinal domain, nouns modified by paucal numerals (2–4) occupy a structured but internally differentiated region, integrating into the broader cardinal plane while exhibiting cohesion distinct from both singular and higher-numeral contexts. By contrast, grammatical case, semantic class, lemma identity, and abstractness contribute only weak and localized effects. These results suggest that numeral-induced morphosyntactic constraints exert a stronger influence on distributional structure than morphologically encoded number alone. More broadly, the findings indicate that contemporary Ukrainian retains distributional traces of historically motivated numeral distinctions, including residual effects associated with the former dual and the paucal range, manifesting as gradient rather than categorical patterns in the embedding space.

• Details
• Contents