Merging Multilingual Information Retrieval Results Based on Prediction of Retrieval Effectiveness

Wen-Cheng Lin, Hsin-Hsi Chen
2004 NTCIR Conference on Evaluation of Information Access Technologies  
This paper deals with Chinese, English and Japanese multilingual information retrieval (MLIR). Merging problem in distributed MLIR is studied. The prediction of retrieval effectiveness is used to determine the merging weight of each intermediate run. The translation penalty and collection weight are considered to improve merging performance. Several merging strategies are experimented. Experimental results show that the performance of normalized-by-top-k merging with translation penalty and
more » ... ection weight is similar to that of raw-score merging and better than that of the other merging strategies.
dblp:conf/ntcir/LinC04 fatcat:zchbfsqa5rembow6nbijd3ufhe