Selecting the n-top retrieval result lists for an effective data fusion

摘要: Although the application of data fusion in information retrieval has yielded good results majority cases, it been noticed that its achievement is dependent on quality input result lists. In order to tackle this problem, paper we explore combination only n-top lists as an alternative all available data. particular, describe a heuristic measure based redundancy and ranking evaluate each list, and, consequently, select presumably n-best per query. Preliminary four IR test collections, containing total 266 queries, employing three different DF methods are encouraging. They indicate proposed approach could significantly outperform achieved by lists, showing improvements mean average precision 10.7%, 3.7% 18.8% when was used along with Maximum RSV, CombMNZ Fuzzy Borda methods.