A holistic and principled approach for the empty-answer problem

Abstract

We propose a principled optimization-based interactive query relaxation framework for queries that return no answers. Given an initial query that returns an empty-answer set, our framework dynamically computes and suggests alternative queries with fewer conditions than those the user has initially requested, in order to help the user arrive at a query with a non-empty-answer, or at a query for which no matter how many additional conditions are ignored, the answer will still be empty. Our proposed approach for suggesting query relaxations is driven by a novel probabilistic framework based on optimizing a wide variety of application-dependent objective functions. We describe optimal and approximate solutions of different optimization problems using the framework. Moreover, we discuss two important extensions to the base framework: the specification of a minimum size on the The work was primarily done while D. Mottin and A. Marascu were at the University of Trento. number of results returned by a relaxed query and the possibility of proposing multiple conditions at the same time. We analyze the proposed solutions, experimentally verify their efficiency and effectiveness, and illustrate their advantages over the existing approaches.

To not miss the forest for the trees -- a holistic approach for explaining missing answers over nested data (extended version)

“…Query refinement and the empty answer problem. Query refinement is also related to our approach [30][31][32]. Query refinements come in two forms: relax queries to return more results or contract queries to return fewer results.…”

Similarity-aware query refinement for data exploration

“…Hence, her constraint is for the query to return a result that contains this specific tuple. Another common example is a user who expects her query to return any result, but the query result is empty [85,91]. Thus, her constraint is for the query to return a non-empty result.…”

“…These two problems aim to refine the original query into a new one, by applying modification operations on the predicates, so that the answer of the new query is likely to contain the tuples that interest the user. In [85,83] they followed an interactive approach to solve this problem, i.e., users are asked for feedback on possible relaxation proposals for their queries. They proposed a probabilistic framework with exact and approximate algorithms which aim to refine a conjunctive query with atomic predicates by dropping some of these predicates (i.e., relaxation sequences) to achieve a non-empty-answer.…”

“…To do that, all relaxation sequences are represented in a tree, rooted by the empty-answer query, and scored by the probability that a user accepts a proposed relaxation. Then, to find the best relaxation sequence, the FullTree [85] algorithm constructs this tree in full and recursively traverses it in a depth-first mode. Since this algorithm is computationally expensive and inappropriate for exploration settings, an optimized algorithm called FastOpt was proposed to minimize the construction cost of this tree by pruning branches of it using lower and upper bounds of the probability scores.…”

“…That is, these techniques provide no guarantee on achieving a target aggregate value, which is the goal of the AQR problem. Further, some of these techniques are limited to only one type of aggregates, i.e., count [90,91], or can only be applied to Boolean databases [85,83].…”