摘要
针对维汉机器翻译中所存在的效率低下以及数据稀疏等问题,本文提出一种多模型融合的词性标注方法。该方法在维吾尔语浅层形态分析的基础上,结合渐进标注模型(Progressive POS,PPOS)对噪音数据的过滤能力及泛化标注模型(Generalize POS,GPOS)的泛化表示能力,对维吾尔语进行词性标注。实验证明,使用该方法进行维吾尔语词性标注,其标注效果已接近实用。
For the Inefficient and data sparse in Uyghur Part-Of-Speech（ POS） tagging,this paper presents a tagging method that merged multi-models. The tagging method based on the shallow morphological analysis of Uyghur,and combined the noise data filtering capability of the Progressive Part-Of-Speech model and generalization capability of the Generalize POS model. Experiments show that with the tagging method mentioned above,the quality of Uyghur Part-Of-Speech tagging is closing to the practical.
出处
《网络新媒体技术》
2014年第1期60-64,共5页
Network New Media Technology
基金
中国科学院战略性先导科技专项课题"新疆少数民族信息处理"(课题编号:XDA06030400)
关键词
维汉机器翻译
维吾尔语词性标注
感知器算法
泛化
复杂形态语言
Uyghur-Chinese machine translation
Part-Of-Speech tagging of Uyghur
perceptron algorithm
generalization
lan-guage with complex morphology
