|M.Sc Student||Izsak Peter|
|Subject||The Search Engines Duel|
|Department||Department of Industrial Engineering and Management||Supervisor||Professor Oren Kurland|
|Full Thesis text|
How can a search engine with a relatively weak relevance ranking function compete with a search engine that has a much stronger (i.e., more effective) function? This question can be the concern, for example, of newly created Web search engines which have to compete with well established engines.
The question just posed can be viewed as a dual challenge which to the best of our knowledge has not been addressed in previous work in the information retrieval research community. One natural way, for the weak engine, to approach the duel is to employ a bi-modal utility function that encompasses effectiveness and diversity with respect to the results of the strong search engine. Specifically, we state as a goal for the weak engine to produce for a query a document result list whose effectiveness (in terms of relevance) does not fall much behind that of the stronger search engine; and, which is quite different than that of the strong engine.
We present a suite of algorithmic responses (i.e., ranking functions) for the weak search engine. These strategies are assumed to get as input the results of the strong search engine for a given query. The strategies are based on pseudo-feedback-based relevance modeling of the results of the strong search engine. Empirical evaluation demonstrates the clear merits of several of the proposed response strategies.