【摘要】 句子相似度的计算在自然语言处理的各个领域有很广泛的应用,但跨语言的句子相似度计算方法却非常少。文中提出一种基于互译特征词对匹配,构建老-汉双语句子相似度计算方法,改进了传统的依赖于词形词序通过计算相同词个数和共有单词的位置信息的相似度计算方法,充分考虑了老挝语和汉语句子中的词汇互译信息、相似概率,避免了由于特征词位置导致的精度丢失。此方法用来最终识别相似度较高的老-汉双语平行句对,依据相似度对源句子和目标句子进行对齐,在老-汉双语平行语料库的建设中使用。实验结果表明,此方法在一定程度上提高了老-汉双语句子相似度计算的准确率。更多还原

【Abstract】 Sentence similarity computation has been widely applied to various fields of natural language processing,but very few cross-language sentence similarity computation methods are available. The similarity calculation method for Laos-Chinese bilingual sentences is constructed based on the reciprocal translated feature word pair matching,by which the traditional similarity calculation method calculating the number of same words and the location information of common words depended on the word form and word order is improved. The lexical mutual translation information and similarity probability in Lao and Chinese sentences are fully considered in this method to avoid the precision loss caused by the location of feature words. This method is used to identify the Laos-Chinese bilingual parallel sentence pairs with higher similarity and align the source sentences with the target sentences according to the similarity. It is used in the construction of the Laos-Chinese bilingual parallel corpus.The experimental results show that this method can improve the accuracy rate of similarity calculation of Laos-Chinese bilingual sentences to some extent.更多还原