This repository hosts the code of the algorithm implemented for our Master Thesis. We called it "EXPLORER" (EXPLanation Oriented query Reverse EngineeRing) since it aims at finding explanations of query answers using query reverse engineering
The following is the Thesis Abstract
The aim of this work is to find ways to discover hidden relations among a set of data extracted from a database. This relation enquiry is helpful to understand if those extracted data have some distinctive characteristics and in particular to highlight if some unethical principles are present. Querying data from databases is very simple and fast. Database management systems (DBMS) allow users to extract data in a very efficient way. They are optimized to lower the latency of a data request and they are designed to allow data security and consistency. Given a query, the database management system analyses data, computes the result and delivers it to the user. Once the user has the result, s/he could be interested in knowing if some valuable relation is present among the elements of the result. In particular, it could be very useful to know if those tuples are related in some way or if some unethical principles, with respect to the rest of the database, are injected in the result. To perform a similar analysis, the user cannot rely on the DBMS since, typically, they do not offer such instruments.
We have then decided to develop a tool, called EXPLORER, for EXPLanation Oriented query Reverse EngineeRing, that supports an analysis on sensitive aspects of the data extraction. The input to be given is a group of tuples that could be the result of some known or unknown query. Starting from the result, EXPLORER uses different query reverse engineering methods we have implemented in order to find equivalent queries. This set of equivalent queries highlights the other ways in which the input result can be produced. Following this process, the implicit relations that tuples in the result have can be easily pointed out. These implicit relations, if put together, can produce an explanation. An explanation, as the word says, explains how the result has been obtained or, in other words, why the result is made up of those specific tuples. The natural explanation of a result is of course the query that has initially generated it. However, it can be very valuable to analyse the equivalent queries to know if ethically sensitive relations on data are present.
This explanations-providing tool can also be thought as an extension of a DBMS that offers an analysis on query results every time the user is interested in having it. Moreover, the tool can also be used as a simple query reverse engineering method to infer equivalent queries. This allows our method to be useful for several purposes.