Abstract (EN)

Leveraging query logs benefits the users analyzing large data warehouses with OLAP queries. But so far nothing exists to allow the user to have concise and usable representations of what is in the log. In this article, we present a framework for summarizing and querying OLAP query logs. The basic idea is that a query summarizes another query and that a log, which is a sequence of queries, summarizes another log. Our formal framework includes a language to declaratively specify a summary, and a language for querying and manipulating logs. We also propose a simple measure based on precision and recall, to assess the quality of summaries, and two strategies for automatically computing log summaries of good quality. Finally we show how some simple properties on the summaries can be used to query the log efficiently. The framework is implemented using the Mondrian open source OLAP engine. Its interest is illustrated with experiments on synthetic yet realistic MDX query logs.