Paraphrasing of Mythological Discourse in Uzbek: Toward Style-Aware and Symbol-Sensitive NLP Models

This electronic document is a "live" template and already defines the components of your paper [title, text, heads, etc.] in its style sheet. *CRITICAL: Do Not Use Symbols, Special Characters, Footnotes, or Math in Paper Title or Abstract. (Abstract) This study proposes a style-aware paraphrasing approach for the Uzbek language, with a focus on mythological and scientific discourse. Uzbek, a low-resource and morphologically rich language, poses specific challenges for paraphrase generation, especially across functional styles. We present a hybrid system combining rule-based methods with fine-tuned multilingual transformer models (mBART, T5) trained on a custom-built parallel corpus of Uzbek texts. Our evaluation demonstrates that transformer-based models significantly outperform rule-based baselines in terms of fluency, semantic preservation, and stylistic accuracy. This research contributes to low-resource NLP and opens new directions for computationally processing culturally embedded narratives in Central Asian languages.

Перевод пока недоступен