维吾尔语-乌兹别克语机器翻译研究

发布时间：2018-05-07 09:20

本文选题：机器翻译 + 乌兹别克语　；参考：《新疆大学》2017年硕士论文

【摘要】：自从计算机发明以来,人们对机器翻译的兴趣越来越大。机器翻译是指计算机把一种自然语言(源语言)转换成另一种自然语言(目标语言)的过程。维吾尔语和乌兹别克语在单词结构、构形、词性、句法等方面有很多相同点或相似之处。因此,实现维吾尔语和乌兹别克语的机器翻译比较容易。本文中提出了基于词典和规则相结合的维吾尔语和乌兹别克语机器翻译方法。首先分别研究维吾尔语和乌兹别克语的字母及其对应关系、音节划分规则和音节关系、单词结构和词类关系、句子结构和成分关系等。在此基础上,构建两种语言对齐词典和字母对应规则库并实现基于词典的维吾尔语和乌兹别克语翻译方法和基于字母转换规则的维吾尔语和乌兹别克语翻译方法;随后结合两种翻译方法实现基于词典和规则相结合的维吾尔语和乌兹别克语机器翻译系统。乌兹别克语到维吾尔语方向的翻译中发现,乌兹别克语字母和维吾尔语字母之间存在歧义现象,一个乌兹别克语字母对应两个及以上维吾尔语字母。对这种歧义现象用最大熵特征选择方法来判断歧义字母对应的维吾尔语字母并生成目标语言。初步优化了基于词典和规则相结合的乌兹别克语-维吾尔语机器翻译系统。小规模的实验结果表明该系统在维吾尔语和乌兹别克语机器翻译准确率较高,其中乌兹别克语到维吾尔语方向的翻译质量比维吾尔语到乌兹别克语方向的翻译质量高。在以后的工作中,该系统还可通过双语词干库和词缀库规模的不断扩大、双语翻译规则的不断完善,以及可以用现有的词典和规则相结合的翻译基础上分析词性变化等方法有效提高翻译质量。
[Abstract]:Since the invention of computer, people have become more and more interested in machine translation. Machine translation is the process by which a computer converts one natural language (source language) into another (target language). Uygur and Uzbek have many similarities or similarities in word structure, configuration, part of speech, syntax and so on. Therefore, the implementation of Uighur and Uzbek machine translation is easier. In this paper, a method of Uygur and Uzbek machine translation based on dictionary and rule is proposed. This paper first studies the letters and their corresponding relations between Uyghur and Uzbek, the syllable division rule and syllable relation, the word structure and part of speech relationship, the sentence structure and the component relation and so on. On the basis of this, we construct the dictionary and alphabetic correspondence rule library and realize the translation methods of Uygur and Uzbek languages based on dictionaries and the translation methods of Uyghur and Uzbek languages based on alphabetic conversion rules. Then the Uygur and Uzbek machine translation systems based on dictionaries and rules are implemented in combination with two translation methods. In the translation from Uzbek to Uygur, there is ambiguity between the Uzbek and Uygur letters, and one Uzbek letter corresponds to two or more Uygur letters. The maximum entropy feature selection method is used to determine the Uygur letters corresponding to the ambiguous letters and to generate the target language. The Uzbek-Uygur machine translation system based on dictionaries and rules is preliminarily optimized. The results of small-scale experiments show that the accuracy of the system in Uygur and Uzbek machine translation is higher, and the translation quality from Uzbek to Uygur is higher than that from Uygur to Uzbek. In the future work, the system can also expand the scale of bilingual stem database and affix database, and improve the rules of bilingual translation. The translation quality can be improved by analyzing the change of part of speech on the basis of the combination of existing dictionaries and rules.
【学位授予单位】：新疆大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.2

【参考文献】