Asian Language Parsing Evaluated by Hummingbird SearchServer TM at NTCIR-3.
Stephen Tomlinson · NTCIR · 2002
Hummingbird submitted ranked result sets for the Chinese, Japanese and Korean Single Language In- formation Retrieval tracks of the Cross-Language Re- trieval Task of the 3rd NII-NACSIS Test Collection for IR Systems Workshop (NTCIR-3). SearchServer 5.3's segmenter for Asian text, compared to an overlap- ping n-gram approach, was found to modestly in- crease precision scores for Japanese, to have a neu- tral impact for Chinese, and to be detrimental for Ko- rean. SearchServer's option to case normalize Hira- gana and Katakana n-grams increased precision sub- stantially for one Japanese query and was of neutral impact for the others. Newline suppression was found to be of only minor benefit for n-gram parsing. Nor- malizing Han characters to Hangul had almost no ef- fect on the Korean test collection. Keywords: Hum- mingbird, SearchServer, NTCIR, Chinese, Japanese, Korean, Asian, parsing, n-grams, segmenting.