How Similar are Chinese and Japanese for Cross-Language Information Retrieval?

the Fifth NTCIR Workshop Meeting on Evaluation of Information Access Technologies: Information Retrieval, Question Answering and Cross-Lingual Information Access

计算机科学;图书情报档案学

Fredric C. Gey ;

Others  :  http://research.nii.ac.jp/ntcir/workshop/OnlineProceedings5/data/CLIR/NTCIR5-CLIR-GeyF.pdf
PID  :  52316

来源: CEUR

【 摘 要 】

For NTCIR Workshop 5 UC Berkeleyparticipated in the bilingual task of the CLIRtrack.Our focus was on Chinese topicsearches against the Japanese News documentcollection, and on Japanese topic searchagainst the Chinese News DocumentCollection. Extending our work of NTCIR 4workshop, we performed search experiments tosegment and use Chinese search topics directlyas if they were Japanese topics and vice versa. We also utilized a commercial MachineTranslation (MT) between the two languages,with English as a pivot language. The bestperformance of Chinese topic search forJapanese documents was achieved using ahybrid approach which combined MT pivottranslation with direct use of Chinese topicexpressions.