In this paper, the de+nition of a Meta search engine model named SOFIA(SOft Fusion of Information Access) is proposed that applies a soft and -exible fusion of the ranked lists of documents retrieved by distinct search engines available over the Internet. The peculiarity of the fusion is that the +nal rank is not determined by a linear combination of the ranks in the lists, as generally happens in most metasearch engines. Instead, a linguistic quanti+er modeled by an Induced Ordered Average (IOWA) operator that allows us to realize soft fusions in between that of the intersection and the union of the lists respectively expresses the fusion criterion. Flexibility is obtained by allowing user to specify his/her retrieval attitude that can be either recall or precision oriented, and by computing distinct +tness scores of the search engines based on a relevance feedback mechanism. These scores are used to de+ne the IOWAreorder vector. In this way, the search engines with highest +tness determine more heavily the ranking of the documents in the fused list.
A model for a Soft Fusion of Information Accesses on the web.
Bordogna G;
2004
Abstract
In this paper, the de+nition of a Meta search engine model named SOFIA(SOft Fusion of Information Access) is proposed that applies a soft and -exible fusion of the ranked lists of documents retrieved by distinct search engines available over the Internet. The peculiarity of the fusion is that the +nal rank is not determined by a linear combination of the ranks in the lists, as generally happens in most metasearch engines. Instead, a linguistic quanti+er modeled by an Induced Ordered Average (IOWA) operator that allows us to realize soft fusions in between that of the intersection and the union of the lists respectively expresses the fusion criterion. Flexibility is obtained by allowing user to specify his/her retrieval attitude that can be either recall or precision oriented, and by computing distinct +tness scores of the search engines based on a relevance feedback mechanism. These scores are used to de+ne the IOWAreorder vector. In this way, the search engines with highest +tness determine more heavily the ranking of the documents in the fused list.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
