A Linguistically Grounded Sense Inventory for Uzbek Polysemous Words
Year: 2026
Type: conference-paper
Abstract: This resource-construction study presents a linguistically grounded sense inventory for Uzbek polysemous words. Each lexicographic sense record is represented along two independent dimensions: a semantic-shift label linking it to a designated prototype and a coarse semantic-domain label indicating the conceptual area in which it is interpreted. The working inventory contains 1,092 lemmas and 2,846 senses, with five shift labels and 24 domains. There are 1,091 PRIMARY records (38.3%); one lemma still lacks a resolved prototype and is explicitly flagged for re-audit. Among 1,755 non-PRIMARY records, metaphor and metonymy jointly account for 56.4%, while abstract and social domains cover 47.0% of all senses. Prototype selection uses dictionary order only as an initial cue and then considers semantic autonomy, denotational breadth, concreteness, and explanatory productivity. Borderline cases may retain a secondary shift label, confidence state, and adjudication note. The paper reports a provisional single-expert resource and descriptive audit, not a validated word-sense-disambiguation system. No inter-annotator agreement, corpus-frequency estimate, or downstream performance gain is claimed. A provenance and rights audit plus independent double annotation of a stratified sample are defined as release gates before a gold version.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7224904210