Improving Thai-English Statistical Machine Translation In Filtering Process Using A Self-Organizing Map

摘要：

The purpose of this work was to improve the efficiency of Statistical Machine Translation (SMT) using a Self-Organizing Map (SOM). Normally, ordinary SMT has two major processes, training process and translating process. The training process is the process that statistically prepares resources from a number of bilingual corpuses for being used in the translating process. However, the obtained resources still contain many unrelated data. This research mainly focuses on a new method that using a SOM to filter only unrelated data out the final translation model as much as possible. The initial results show that using a SOM in filtering process is capable of screening out incorrect pairings more effective than normal statistical approach. Therefore, the better statistical translation model can be created. The project assumption is that the efficiency of ThaiEnglish SMT could be improved by using this improved statistical translation model. Preliminary results show promising improvement in filtering process.

关键词： Self-Organizing Map Statistical Machine Translation Data Mining

作者: Singha Wongdeethai Jumpol polvichai

作者单位: Department of Computer Engineering King Mongkuts University of Technology Thonburi Bangkok, Thailand

会议类型: 国际会议

会议名称: 2011 International Conference on Database and Data Mining(ICDDM 2011)(2011年数据库和数据挖掘国际会议)

会议地点: 三亚

会议语种:英文

页码: 162-165

在线出版日期: 2011-03-25（万方平台首次上网日期，不代表论文的发表时间）