Results processing in a heterogeneous word

Abstract

Distributed digital libraries allow users to access data of different modalities, from different information sources, and ranked by different criteria. Most applications make too many assumptions, and need too much information. We assume that each information retrieval model is satisfactory in its own context. Based on this assumption, we propose two results processing methods: Ranking by Sources (RBS) and Simply Merging Results (SMR). In RBS, we define satisfied ranking, which is the ranking satisfying most source rankings, and satisfied distance, which indicates how a specific source ranking suits the satisfied ranking. RBS groups the results by the ranked sources, which is sorted by their satisfied distances. In SMR, for each result, we substitute the normalized score for its original scores, and then merge them using normalized scores. The experiment showed that our methods are very feasible in the rapid expanding distributed digital libraries.