One of the main difficulties in Chinese-Korean cross-language information retrieval is to translate named entities (NE) in queries. Unlike common words, most NE's are not found in bilingual dictionaries. This paper presents a pattern-based method of finding NE translations online. The most important feature of our system is that patterns are generated and weighed automatically, saving considerable human effort. Our experimental data consists of 160 Chinese-Korean NE pairs selected fromWikipedia in five domains. Our approach can achieve a very high MAP of 0.84, which demonstrates our system's practicability.
|State||Published - 2008|
|Event||20th Conference on Computational Linguistics and Speech Processing, ROCLING 2008 - Taipei, Taiwan|
Duration: 4 Sep 2008 → 5 Sep 2008
|Conference||20th Conference on Computational Linguistics and Speech Processing, ROCLING 2008|
|Period||4/09/08 → 5/09/08|