点击次数：

论文类型：期刊论文

发表时间：2011-03-01

发表刊物：ICIC Express Letters

收录刊物：EI、Scopus

卷号：5

期号：3

页面范围：701-706

ISSN号：1881803X

摘要：The generation technique of inducing translation templates is a key problem for many EBMT systems. The objectives of this research are to improve translation templates and the way inducing them in order to raise work efficiency of the EBMT systems. Translation templates with grammar-semantic typed variables are presented and a set of machine learning methods that induce translation templates with grammar-semantic typed variables is proposed. The grammar-semantic type constraints avoid some of the wrong translations for further restricting the usage of translation templates in certain contexts. The proposed machine learning methods further restrict the replacements of variables and effectively increase the percentage of producing correct translation results. The quality of the presented translation templates and the proposed machine learning methods is measured experimentally. ? 2011 ISSN 1881-803X.