A lattice framework for reusing top-k query results
Brandeis Hill · 2005
In this paper, we address the issue of reusing the top-k query results for information retrieval applications. A query comprises of a number of search terms in which multiple queries use the same search terms. The re-occurring search terms and associated results are indexed and easily accessible for retrieval. The individual results of the search terms are merged during the query processing stage and return the best combined results. Hence, the queries are observed as independent events but should be viewed in tandem. We present a lattice framework that groups frequently requested search terms and corresponding results in a cache. The stored results provide a reduction in database accesses while minimally affecting the performance. We compare the aggregate ranking (from the actual rankings) to the lattice-based aggregate ranking (using at least one stored result). We perform an empirical study of our proposed framework in which we analyze its benefits and drawbacks to the conventional method.