Filling the Gap for Uzbek: Creating Translation Resources for Southern Uzbek

TL;DR

This paper introduces new translation resources and a model for Southern Uzbek, a low-resource Turkic language, including datasets, a fine-tuned model, and a post-processing method to enhance translation quality.

Contribution

It provides the first comprehensive resources and a specialized model for Southern Uzbek, addressing its underrepresentation in NLP.

Findings

New parallel datasets and dev set released.

A fine-tuned NLLB-200 model for Southern Uzbek.

Post-processing improves morphological boundary handling.

Abstract

Southern Uzbek (uzs) is a Turkic language variety spoken by around 5 million people in Afghanistan and differs significantly from Northern Uzbek (uzn) in phonology, lexicon, and orthography. Despite the large number of speakers, Southern Uzbek is underrepresented in natural language processing. We present new resources for Southern Uzbek machine translation, including a 997-sentence FLORES+ dev set, 39,994 parallel sentences from dictionary, literary, and web sources, and a fine-tuned NLLB-200 model (lutfiy). We also propose a post-processing method for restoring Arabic-script half-space characters, which improves handling of morphological boundaries. All datasets, models, and tools are released publicly to support future work on Southern Uzbek and other low-resource languages.