| 摘要 |
<p><P>PROBLEM TO BE SOLVED: To provide a word segmentation system and a word segmentation method, which enhance the quality of translation by integrating a plurality of source language word segmentation systems into SMT (Statistical Machine Translator) decode processing. <P>SOLUTION: The phrase table generator 10 includes: a storage 30 for storing bilingual corpora 32, 34 of translation pairs. Each translation pair includes a source sentence of the first language 34 and a target sentence of the second language 32. The generator further includes a classifier trainer 12 for training a statistical machine translator (SMT) utilizing the corpus. The SMT outputs a phrase table 16 during the training. The generator 10 further includes a phrase table merger 18 for integrating a plurality of phrase tables 16 into an integrated phrase table 20. <P>COPYRIGHT: (C)2011,JPO&INPIT</p> |
