An axiomatic study of Web meta search engines

The meta search engine is an effective technique to improve the quality of retrieval results of Web pages on the internet [26] [32]. It can be modeled as a group decision problem by treating each search engine as a decision-maker and each document as a candidate. From several retrieval results for the same query from different web search engines, a globally combined and enhanced result can be obtained by means of a combination function.; Two methods can be used to find a combination function according to the different representations of results from search engines. Numerical combination deals with specific Retrieval Status Value (RSV) associated with each document. By specifying a weighting vector, a combination function is defined. A qualitative combination uses the qualitative information induced by RSVs of documents. The group decision is a weak order on the document set.; The emphasis of the thesis is to provide a solid foundation of meta search engines, rather than new combination methods. Based on group decision and measurement theories, a unified framework is established in an axiomatic setting for the study and analysis of the combination problem. The combination of multiple web search engines is reduced to finding an effective combination function. Two qualitative models, called ‘outranking preference model’ and ‘distance-based model’, are developed to construct combination functions. The advantages of qualitative methods over numerical ones are presented. The relationship between the existence of combination functions and the axioms satisfied by the corresponding functions is discussed. Within the proposed axiomatic framework, one can easily identify the assumption, consequence, and implication of existing methods. These analytical results can provide the meta search engine designer with an insight into the combination process.