Evaluation of Japanese Phrasal Indexing with Large Test Collection

Sumio Fujita
2000 Open research Areas in Information Retrieval  
Effect o f phrasal indexing had been considered as uncertain by IR research society but after TR E C evaluation started, many groups adopted phrasal indexing and reported positive results. In spite o f such long controversy, it is not yet clear if phrasal indexing is effective for text retrieval. Furthermore, the problem is more com plicated for languages like Japanese where word boundaries are not marked on written text. We used the first large scale test collection for Japanese IR recently
more » ... eased and carried out experim ents for phrasal indexing and its w eighting issues in view o f various length o f queries. The results show that phrasal indexing outperform s single word only indexing with long queries while single word only indexing perform s slightly better with short queries. Down-weighting m ethod for phrasal terms is also evaluated and perform ance improvement is observed. Another experim ent clearly showed correlation between the length o f queries and effect o f phrasal indexing.
dblp:conf/riao/Fujita00 fatcat:xz3rpqvofbfedlbseqg2uzu2oa