¢¦TOP¢¦³èÆ°¼ÂÀÓ¢¦£²£°£°£³Ç¯¢¦
(37) Effectiveness of Automatic Extraction of Bilingual Collocations Using Recursive Chain-link-type Learning
¡¡¡¡¡¡Proceedings of the Ninth Machine Translation Summit, pp.102-109, 2003-9
This paper describes a new method for automatic extraction of bilingual collocations from a parallel corpus. We use Recursive Chain-link-type Learning (RCL), which is a learning algorithm, to extract bilingual collocations. In our method, new
bilingual collocations are extracted efficiently from sentence pairs of source and target languages using only character strings of the bilingual collocations extracted previously. Therefore, a system using RCL can extract bilingual collocations with no need for analytical knowledge, even if the frequency of appearance is very low. Evaluation experiment results show that a system using RCL efficiently extracts bilingual collocations which correspond to English and Japanese word pairs. The recall of correct translation for unknown words in the commercial English-to-Japanese machine translation system was
72.4% when frequency of appearance of the correct translations is only one or two in a parallel corpus. This result shows hat our method very efficiently extracts bilingual collocations.
>>