Automatic assessment of interactive OLAP explorations

Abstract

Interactive Database Exploration (IDE) is the process of exploring a database by means of a sequence of queries aiming at answering an often imprecise user information need. In this paper, we are interested in the following problem: how to automatically assess the quality of such an exploration. We study this problem under the following angles. First, we formulate the hypothesis that the quality of the exploration can be measured by evaluating the improvement of the skill of writing queries that contribute to the exploration. Second, we restrict to a particular use case of database exploration, namely OLAP explorations of data cubes. Third, we propose to use simple query features to model its contribution to an exploration. The first hypothesis allows to use the Knowledge Tracing, a popular model for skill acquisition, to measure the evolution of the ability to write contributive queries. The restriction to OLAP exploration allows to take advantage of well known OLAP primitives and schema. Finally, using query features allows to apply a supervised learning approach to model query contribution. We show on both real and artificial explorations that automatic assessment of OLAP explorations is feasible and is consistent with the user’s and expert’s viewpoints.