Abstract: Aiming at bilingual corpora is critical resources for developing statistical machine translation system, this paper presents a method which automatically mines bilingual parallel Web page form Web. Different from mining data from pre-specified Web sites, the system is developed to mine parallel Web page from the entire Web, it is greatly suitable for new content domains and language pairs. It implements a parallel Web page mining system. Experimental results show that the system can provide large scale and high quality parallel Web page for statistical machine translation.

Key words: natural language processing, statistical machine translation, bilingual corpora, Web mining

摘要： 针对双语语料是开发统计机器翻译系统的重要资源，提出一种从网络中自动挖掘双语平行网页的方法。与传统从指定网站中挖掘平行网页的方法不同，该方法从整个互联网中自动挖掘平行网页，对新的语言对和内容领域有很强的适应能力，实现双语平行网页挖掘的系统。实验结果显示，该系统可以为统计机器翻译系统提供大量高质量的平行网页。

关键词: 自然语言处理, 统计机器翻译, 双语语料, 网络挖掘

CLC Number: 

• TP312