汉语手语翻译系统研究与实现

发布时间：2018-09-18 09:20

【摘要】：本文研究了汉语与手语(Sigh Language)这两种不同语言形式各自的特点,并结合语言学的知识加以说明。另外,本论文根据汉语到手语翻译的理论基础,探讨汉语与手语之间的对应关系以及实现形式,该研究成果可以方便人们学习和认识手语,并能够在手语的翻译领域得到广泛应用,促进人类对自然语言的认知和学习。从语言学的角度来说,汉语的“基本词”是汉字。中国手语也有基本词的概念,但是《中国手语》一书却是以词为单位进行描述,没有提及手语的构词法,为了挖掘手语词汇中手语的构词规则,本文通过DTW距离去衡量手语词汇和基本词的关系,并提出手语词汇编码的思想,将手语词汇表现成基本词的组合。在中文的信息处理方面,能否正确的分词是计算机能否理解语义的关键,而理解句子的语义是正确翻译句子的前提。所以在中文手语的翻译过程中,有必要对中文句子进行分词处理。目前常用的中文分词方法大致有三种:基于最大匹配的方法、基于大规模语料库的统计学方法和基于理解的方法。本文通过对这三种分词方法的研究,总结出各种方法的优缺点。为了充分发挥各种分词方法的优点,本文采用双层哈希字典并结合改进的正向匹配算法,对中文进行分词处理,同时判断句中是否存在交集型歧义,并采用全切分算法和统计学方法相结合的方法处理交集型歧义,对于句中是否存在组合型歧义,本文采用基于统计学的算法找出歧义字段,并依据DSM算法对组合型算法进行处理。本论文结合前人对手语和汉语的研究基础,对传统的中文处理方法进行总结、改进,并对手语的语言规律进行研究总结,实现汉语向手语的转换,实现了通过计算机进行汉语到手语翻译的功能,达到了软件设计说明书的要求。
[Abstract]:This paper studies the characteristics of Chinese and sign language (Sigh Language), and explains them with linguistic knowledge. In addition, according to the theoretical basis of Chinese to hand language translation, this thesis discusses the corresponding relationship between Chinese and sign language and its realization form. The research results can facilitate people to learn and understand sign language. And it can be widely used in the field of sign language translation to promote human cognition and learning of natural language. From the linguistic point of view, the basic words in Chinese are Chinese characters. Chinese sign language also has the concept of basic words, but the book "Chinese sign language" is described in terms of words, without any mention of sign language word-formation, in order to excavate the sign language word-formation rules in sign language vocabulary. In this paper, the relationship between sign language vocabulary and basic words is measured by DTW distance, and the idea of sign language lexical coding is put forward to combine the ready-made basic words of sign language vocabulary. In the aspect of Chinese information processing, whether the word segmentation is correct is the key to the computer's understanding of the semantics, and the understanding of the sentence semantics is the premise of the correct translation of the sentence. Therefore, in the process of Chinese sign language translation, it is necessary to deal with Chinese sentence segmentation. There are three commonly used Chinese word segmentation methods at present: one based on maximum matching, one based on large-scale corpus and one based on understanding. This paper summarizes the advantages and disadvantages of these three word segmentation methods. In order to give full play to the advantages of various word segmentation methods, this paper uses a double-layer hash dictionary combined with an improved forward matching algorithm to deal with Chinese word segmentation, and at the same time to determine whether there is an intersection of ambiguity in sentences. And the method of combining the total segmentation algorithm with the statistical method is used to deal with the intersection ambiguity. As to whether there is combinatorial ambiguity in the sentence, this paper uses the algorithm based on statistics to find out the ambiguous fields. And according to the DSM algorithm to deal with the combined algorithm. Based on the previous researches on sign language and Chinese, this thesis summarizes and improves the traditional Chinese processing methods, and summarizes the language rules of sign language to realize the conversion from Chinese to sign language. The function of translating Chinese to hand language by computer is realized, and the requirement of software design instruction is achieved.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.2

【参考文献】