Joining the results of heterogeneous search engines

摘要

In recent years, while search engines have become more and more powerful, several specialized search engines have been developed for different domains (e.g. library services, services dedicated to specific business sectors, geographic services, and so on). While such services beat generic search engines in their specific domain, they do not enable cross-references; therefore, they are of little use when queries require input from two or more of such services (e.g., “find papers in VLDB 2000 authored by a member of a specified department” or “books sold online written by prolific database authors” or “vegetarian restaurants in the surroundings of San Francisco”). In this paper, we study how to join heterogeneous search engines and get a unique answer that satisfies conjunctive queries, where each query can be routed to a specialized engine. The paper includes both the theoretical framework for stating such problem and the description of pragmatic solutions based on web service technology. We present several algorithms for the efficient computation of join results under several cost model assumptions.