Mirror mirror on the wall, which query's fairest of them all?
Georgia Koutrika, Alkis Simitsis · Conference on Innovative Data Systems Research · 2013
Scientists heavily use database technologies to store and analyze massive amounts of experimental data. However these tasks require (sometimes, advanced) knowledge of a query language, which is not necessarily amongst the normal skills of a scientist. Database experts may help in the initial construction of such queries, which are then stored in a repository. Afterwards, instead of writing queries from scratch, stored queries can be re-used with few or no changes by people in the same or a dierent research group. However, as much as this is a desired reality, we are actually stepping into dreamland. Searching a pool of database queries for those that t in a certain analysis is not easy even for experienced database users. One solution could be to provide a description for each query and then, use a keyword search technique to identify queries of interest. But doing this manually is time consuming, and its success relies on the uniformity of the descriptions provided {which is extremely hard when many query writers are involved in this eort as typically happens. In this paper, we envision a system that can understand the problems past queries solve, the computations they involve and their signicance, and can help the user compose the queries required for the task at hand using this knowledge from past queries.