摘要
在信息检索,文本挖掘以及基于实例的机器翻译中,相似度计算都是一个关键问题。在实例机器翻译中,相似度计算一般是基于字符、词的匹配以及向量空间模型,但基于句子语义结构的相似度研究还不多见。借助了汉语框架语义网(Chinese FrameNet,简称CFN)的场景语义描述优势,提出了一种新的面向EBMT进行实例相似度计算的方法。该方法主要基于CFN从句子整体结构相似和各语义块内部相似两个角度来度量句子相似度,将这两部分的相似度结果进行凸组合作为待翻译句子与候选实例之间的相似度值。实验结果表明,与传统方法相比,所提出的这种方法是有效的。
In the information retrieval,text mining,as well as Example based Machine Translation,Similarity calculation is a key issue.In the Example based Machine Translation,General similarity calculation is based on the characters,word matching,and vector space model.However,the study of the similarity based on the semantic structure of sentences is still rare.In this paper, with the semantic description advantage of Chinese FrameNet,We proposed a new method of similarity calculation oriented EBMT.This method is mainly based on the CFN from the overall structure of the sentence and the internal of the semantic block to measure the similarity between two sentences,then the convex combination of the results of these two Similarities is considered to be the similarity between Sentence to be translated and the candidate example.The experimental results show that compared with traditional methods the method proposed in this paper is effective.
出处
《电脑开发与应用》
2011年第6期58-60,共3页
Computer Development & Applications
关键词
相似度
实例机器翻译
汉语框架网
框架语义
similarity
example based machine translation
chinese frameNet
frame semantic
