The effects of separate and merged indexes and word normalization in multilingual CLIR
Airio, Eija (2005)
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
Multilingual IR may be performed in two environments: there may exist a separate index for each target language, or all the languages may be indexed in a merged index. In the first case, retrieval must be performed separately in each index, after which the result lists have to be merged. In the case of the merged index, there are two alternatives: either to perform retrieval with a merged query (all the languages in the same query), or to perform distinct retrievals in each language, and merge the result lists. Further, there are several indexing approaches concerning word normalization. The present paper examines the impact of stemming compared with inflected retrieval in multilingual IR when there are separate indexes / a merged index. Four different result list merging approaches are compared with each other. The best result was achieved when retrieval was performed in separate indexes and result lists were merged. Stemming seems to improve the results compared with inflected retrieval.