What question did this study set out to answer?

This research aims to evaluate the performance of various multilingual neural machine translation models for the Uzbek-Russian language pair after fine-tuning.

June 3, 2026SCIENCE TIME0 citationsOpen Access

Сопоставительный Анализ Дообучения Мультиязычных Нейросетевых Моделей Машинного Перевода Для Узбекско-Русской Языковой Пары

АСАвезов Сухроб Собирович

Key Points

This research aims to evaluate the performance of various multilingual neural machine translation models for the Uzbek-Russian language pair after fine-tuning.
Conducted a comparative analysis of six multilingual machine translation models: NLLB-200, M2M-100, mBART-50, OPUS-MT, Google Neural MT, and SeamlessM4T.
Utilized a parallel corpus of 185,000 aligned segments for fine-tuning the models.
Translation quality was assessed using BLEU, chrF, and TER metrics.
NLLB-200-3.3B model showed the highest improvement in translation quality with a BLEU score of 31.6 for Uzbek to Russian.
Identified specific error types for each model, providing insights into performance weaknesses.
Recommendations for model architecture and fine-tuning strategies were made for low-resource Turkic languages.

Abstract

В статье представлен сопоставительный анализ шести мультиязычных моделей нейронного машинного перевода (NLLB-200, M2M-100, mBART-50, OPUS-MT, Google Neural MT, SeamlessM4T) применительно к узбекско-русской языковой паре. Описан процесс дообучения (fine-tuning) указанных моделей на параллельном корпусе объёмом 185 тыс. выровненных сегментов. Качество перевода оценивалось по метрикам BLEU, chrF и TER. Результаты экспериментов показали, что наибольший прирост качества после дообучения демонстрирует модель NLLB-200-3.3B, достигшая значения BLEU 31,6 в направлении узбекский-русский. Обсуждены типичные ошибки каждой модели, предложены рекомендации по выбору архитектуры и стратегии дообучения для низкоресурсных тюркских языков.

Ask AI

Helpful

Bookmark

View Full Paper

Cite This Study

Авезов Сухроб Собирович (2026) studied this question.

synapsesocial.com/papers/6a1fc3a6dee9eb8c0dce5188 https://doi.org/10.66345/stj.v4i3/2.5233

Ask AI

Helpful

Bookmark

View Full Paper

June 3, 2026SCIENCE TIME0 citationsOpen Access

Сопоставительный Анализ Дообучения Мультиязычных Нейросетевых Моделей Машинного Перевода Для Узбекско-Русской Языковой Пары

АСАвезов Сухроб Собирович

Key Points

This research aims to evaluate the performance of various multilingual neural machine translation models for the Uzbek-Russian language pair after fine-tuning.
Conducted a comparative analysis of six multilingual machine translation models: NLLB-200, M2M-100, mBART-50, OPUS-MT, Google Neural MT, and SeamlessM4T.
Utilized a parallel corpus of 185,000 aligned segments for fine-tuning the models.
Translation quality was assessed using BLEU, chrF, and TER metrics.
NLLB-200-3.3B model showed the highest improvement in translation quality with a BLEU score of 31.6 for Uzbek to Russian.
Identified specific error types for each model, providing insights into performance weaknesses.
Recommendations for model architecture and fine-tuning strategies were made for low-resource Turkic languages.

Abstract

Ask AI

Helpful

Bookmark

View Full Paper

Cite This Study

Авезов Сухроб Собирович (2026) studied this question.

synapsesocial.com/papers/6a1fc3a6dee9eb8c0dce5188 https://doi.org/10.66345/stj.v4i3/2.5233

Ask AI

Helpful

Bookmark

View Full Paper