Modeling Uzbek Paraphrase Structures through Travel Discourse Analysis
Abstract
This study proposes a style-aware paraphrasing approach for the Uzbek language by leveraging classical and modern travel literature as a foundational corpus. Recognizing the critical need for paraphrasing tools in low-resource language contexts, the research introduces a tri-parallel corpus built from aligned narrative, journalistic, and academic variants of Uzbek travel texts, including Safarnoma by Nosir Khisraw and Turkistan Solo by Ella Maillart. Through this corpus, we train a paraphrasing model capable of performing register-sensitive transformations while preserving semantic fidelity and stylistic integrity. The study demonstrates how travel narratives serve as both linguistic and cultural data, offering rich idiomatic, syntactic, and stylistic features that support intelligent style transfer. Our model supports both rule-based and neural sequence-to-sequence frameworks, with potential applications in education, journalism, digital humanities, and Uzbek language modernization. The results affirm the importance of culturally grounded corpora in building NLP tools for underrepresented languages and present a scalable framework for similar efforts in other linguistic contexts.