Skip to content
DSML.KZ
Publications

Talk · International Conference on Computer Science and Engineering · September 1, 2026

Linguistic Status of the Subject in Uzbek Legal Discourse and Its Implications for Natural Language Processing

Khamroeva Shahlo Mirdjonovna, Xolida Ubaydullayeva, Gulnoza Salimova, Raximova Norbibish Sharofovna

Gulnoza Salimova
Open via DOI

Abstract

In Uzbek legal discourse, the grammatical subject and the subject of law are often assumed to coincide, yet this assumption has rarely been examined using Uzbek material. This paper investigates how legal subjecthood is morphosyntactically encoded in normative texts and what this means for automatic processing of Uzbek legal language. Using a mixed-methods approach combining qualitative discourse analysis with quantitative distributional analysis, we examined 187 clauses from five Uzbek normative acts, including the Constitution, the Civil Code, the Criminal Code, the Labor Code, and the Code of Administrative Responsibility. Four encoding strategies were identified: an overt nominative subject (Strategy 1, 48.1% of clauses), an anaphorically recoverable subject across coordinated clauses (Strategy 2, 6.4%), a genitive-marked nominal demoted from subject position (Strategy 3, 5.9%), and a passive construction that backgrounds the agent of the legal relation (Strategy 4, 39.6%, consisting of Strategy 4a at 14.4% and Strategy 4b at 25.1%). A chi-square test confirms that the distribution of strategies across obligation versus liability/rights-protection contexts is statistically significant (χ2= 17.91, df = 3, p < 0.001, Cramér's V = 0.309), demonstrating that the grammatical subject and the subject of law coincide reliably in duty-assignment contexts and diverge systematically in liability and rights-protection contexts. To guide future computational work, we provide a detailed error-analysis framework mapping each encoding strategy to predicted failure modes in named entity recognition, dependency parsing, semantic role labeling, and coreference resolution. The paper argues that processing Uzbek legal text requires a pipeline combining syntactic, semantic, and ontology-based components rather than a single annotation layer.

Syntax, Semantics, Linguistic VariationNatural Language Processing TechniquesLegal Language and Interpretationgrammatical subjectsubject of lawlegal subjecthood

More by Gulnoza Salimova