句子相似度的动态规划求解及改进

Calculating the Similarity Degree of Two Sentences by Ameliorated Dynamic Programming

基于例子的机器翻译,其很关键的步骤之一就是如何从语料库中找到待译句子的最佳相似句。论文针对这个问题提出了利用动态规划方法基于句子相似矩阵进行求解的方法。根据这个方法就可以从语料库中为待译句子找到最佳相似句,同时在求解过程中还做了一些改进:利用矩阵分块求解的方法保留了句子的连续相似块,保证了结果的质量,对提高EBMT系统的翻译质量起到了一定的促进作用。

It's a very import step of EBMT to find the best matched example from the example library for the sentence to be translated.In this paper,a reasonable method of dynamic programming is brought out to solve this problem.And using this method,you can find the best matched example for a sentence.At the same time ,it also makes an improvement to the dynamic programming.In order to keep the serial similar block intact,it takes the measure of dividing the similarity matrix into some small matrixes,then calculates the similarity degree of the two sentence.This method will contribute to improving the quality of results of the Example-Based machine translate system.