汉维科技文本的机器翻译质量评估

Machine Translation Quality Evaluation of Chinese-Uyghur Scientific and Technical Texts

摘要: 随着社会的快速发展，科技水平的不断提升，汉维翻译平台也日益增多，这些平台的翻译质量却各不相同。由机器翻译的文本质量是最能体现机器翻译标准程度的依据。并且对维吾尔语使用者在选择机器翻译平台方面有较大帮助。本研究采用200句科技文本为测试语料，使用必应翻译、国语助手、izdax、DeepSeek、谷歌、小牛翻译六个翻译平台导出汉翻维，维翻汉的翻译结果进行测评，对该结果进行了综合打分及对比分析，可供使用者在翻译科技文本时进行参考选择，给机器翻译平台开发者提供优化机器翻译平台的参考依据。研究发现，小牛翻译在汉翻维任务中质量较高，DeepSeek在维翻汉任务中质量位居第一。各翻译平台中出现的主要错误类型有术语错译、逻辑关系处理不当、书写形式不规范等。根据这些错误本文提出了构建领域专用语料库、优化翻译模型架构、建立专有名词音译库、规范书写输出格式等针对性建议。

Abstract: With the rapid development of society and the continuous improvement of technology, the number of Chinese-Uyghur translation platforms is also increasing, but the translation quality of these platforms varies. The quality of machine-translated texts is the most direct indicator of the standard of machine translation and is greatly helpful for Uyghur speakers when choosing a machine translation platform. This study used 200 sentences of scientific and technological texts as test data and evaluated the translation results from Chinese to Uyghur and Uyghur to Chinese generated by six translation platforms: Bing Translator, Guoyu Assistant, izdax, DeepSeek, Google, and Xiaoniu Translator. Comprehensive scoring and comparative analysis were conducted on the results, providing users with a reference when translating scientific and technological texts，and providing developers of machine translation platforms with a reference basis for optimising machine translation platforms. The study found that Xiaoniu Translator achieved higher quality in Chinese to Uyghur translation tasks, while DeepSeek ranked first in Uyghur to Chinese translation tasks. The main types of errors appearing on various translation platforms include incorrect terminology translation, improper handling of logical relationships, and non-standard writing formats. Based on these errors, this paper proposes targeted recommendations such as constructing domain-specific corpora, optimising translation model architectures, establishing a transliteration database for proper nouns, and standardising output writing formats.