Effectiveness of the Learning Method for Partial Correspondence Based on Deletion Possibility in Learning-Type Machine Translation Using Parallel Corpus

Ryo TERASHIMA  Hiroshi ECHIZEN-YA  Kenji ARAKI 

Publication
D - Abstracts of IEICE TRANSACTIONS on Information and Systems (Japanese Edition)  Vol.J93-D  No.3  pp.377-388
Publication Date: 2010/03/01
Online ISSN: 1881-0225
Print ISSN: 1880-4535
Type of Manuscript: PAPER
Category: 
Keyword: 
machine translation,  learning,  partial correspondence,  deletion possibility,  parallel corpus, 

Full Text(in Japanese): PDF(941.9KB)

Summary: 
For machine translations using a parallel corpus, it is effective to determine partial correspondences: pairs of phrases between the source language and the target language in bilingual sentences. In this paper, we propose a new learning method which automatically determines the partial correspondences. In the proposed method, the extraction rules are acquired from the information based on the deletion possibility in each language sentence. Moreover, they possess the information about the first parts (e.g., "a", "the" ) or the last parts in the phrases. Therefore, our method can correctly and efficiently determine the partial correspondences in the bilingual sentences using the acquired extraction rules. From the results of the evaluation experiments, we confirmed that the translation accuracy improved by our proposed method.