FlexIQ: A flexible interactive Querying Framework by Exploiting the Skyline Operator

• We propose a Flexible Interactive Querying (FlexIQ) framework for minimizing the unexpected information as well as maximizing the expected information in the refined query output.
• We provide the definition of data-driven query output semantics which serves as a summarized explanation of the query output.
• We formally define the user feedback and its properties. We also describe approaches to derive minimal user feedback for the purpose of efficient query refinement.
• We then show how one can construct a new boundary for the refined query output by exploiting the skyline operator. We also propose a Trade-Off Algorithm (TOA) for the mentioned query refinement problem.
• We validate our approach with extensive experiments for three different datasets and demonstrate its effectiveness by comparing our results with Decision-Tree based query refinement.

Skyline operator has gained much attention in the last decade and is proved to be valuable for multi-criteria decision making. This paper presents a novel Flexible Interactive Querying (FlexIQ) framework for user feedback-based Select-Project-Join (SPJ) query refinement in databases. In FlexIQ, the user feedback is used to discover the query intent. In addition, we have used the skyline operator to confine the search space of the proposed query refinement algorithms. The user feedback consists of both unexpected information currently present in the query output and expected information that is missing from the query output. Once the feedback is given by the user, our framework refines the initial query by exploiting the skyline operator to minimize the unexpected information as well as maximize the expected information in the refined query output. In our framework, the user can also control different quality metric such as quality of results (e.g., false positive rates, false negative rates and accuracy) and complexity (i.e., quantified as the number of subqueries) in the refined query. We have validated our framework both theoretically and experimentally. In particular, we have demonstrated the effectiveness of our proposed framework by comparing its performance with the naï ve decision tree based query refinement.