Chapter 11: Rank-Join Algorithms for Search Computing

Abstract

Joins represent the basic functional operations of complex query plans in a Search Computing system, as discussed in the previous chapter. In this chapter we provide further insight on this matter, by focusing on algorithms that deal with joining ranked results produced by search services. We cast this problem as a generalization of the traditional rank aggregation problem, i.e., combining several ranked lists of objects to produce a single consensus ranking. Rank-join algorithms, also called top-k join algorithms, aim at determining the best overall results without accessing all the objects. The rank-join problem has been dealt with in the literature by extending rank aggregation algorithms to the case of join in the setting of relational databases. However, previous approaches to top-k queries did not consider some of the distinctive features of search engines on the Web. Indeed, as pointed out in the previous chapter, joins in this context differ from the traditional relational setting for a number of aspects: services can be accessed according to limited patterns, i.e. some inputs need to be provided; accessing services is costly, since they are typically remote; the output is returned in pages of results and typically according to some ranking criterion; multiple search services can be used to answer the same query; users can interact with the system in order to refine their search criteria. This chapter analyzes the challenges that need to be tackled in the design of rank-join algorithms within the context of Search Computing.