面向手语信息处理的维吾尔文本采集的研究  

Research on Uyghur text acquisition for sign language information processing

作　　者：艾山江·亚生 阿里甫·库尔班 张丹丹 Aishanjiang Yasheng;Alifu Kuerban;ZHANG Dandan(School of Information Science and Engineering,Xinjiang University,Urumqi 830046,China;School of Software,Xinjiang University,Urumqi 830008,China)

机构地区：[1]新疆大学信息科学与工程学院,新疆乌鲁木齐830046 [2]新疆大学软件学院,新疆乌鲁木齐830008

出　　处：《现代电子技术》2019年第12期136-139,143,共5页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(61163029);国家自然科学基金资助项目(61562084)~~

摘　　要：从自然语言处理以及深度学习的基本理念、原则出发,为基于中国手语的维吾尔文本信息处理研究提供数据资源,为面向手语信息的手语合成研究,为手语新闻文本编辑研究提供科学依据,对维吾尔文本进行自动分词、自动分句研究,并利用Word2Vec方法进一步建立文本词元库,通过这样保证训练语料的多元化及通用性。最后,利用词干提取方法以及分割字母的方式将维吾尔文本信息转换成手语文本信息,并在此基础上初步采集面向手语信息处理的维吾尔文本信息。实验表明该研究达到了预期目标。Proceeding from the basic concepts and principles of natural language processing and deep learning,research on automatic word and sentence segmentation is conducted for the Uyghur text,so as to provide data resources for Uyghur text information processing research based on Chinese sign language,and scientific basis for research on sign language synthesis and sign language news text editing based on sign language information. The Word2Vec method is used to further establish the text word element library,so as to ensure the diversity and universality of training corpus. The Uyghur text information is converted into the sign language text information by using the methods of word stem extraction and letter segmentation. On this basis,the Uyghur text information based on sign language processing is initially collected. The experimental results show that the research can achieve the expected research goal.

关 键 词：深度学习 Word2Vec方法 手语合成 文本词元库 词干提取 字母分割 

分 类 号：TN912.34.34[电子电信—通信与信息系统] TP311.1[电子电信—信息与通信工程]