Query Similarity for Approximate Query Answering

Query rewriting in heterogeneous environments assumes mappings that are complete. In reality and especially in the Big Data era it is rarely the case that such complete sets of mappings exist between sources, and the presence of partial mappings is the norm rather than the exception. So, practically, existing rewriting algorithms fail in the majority of cases. The solution is to approximate original queries with others that can be answered by existing mappings. Approximate queries bear some similarity to original ones in terms of structure and semantics. In this paper we investigate the notion of such query similarity and we introduce the use of query similarity functions to this end. We also present a methodology for the construction of such functions. We employ exemplary similarity functions created with the proposed methodology into recent algorithms for approximate query answering and show experimental results for the influence of the similarity function to the efficiency of the algorithms.