This repository provides an implementation of the framework for answering why-not-yet questions over top-k queries. Also included are code and data to reproduce the experiments from our VLDB 2023 paper: Why not yet: Fixing a top-k ranking that is not fair to individuals.
The source code of why-not-yet is written in Java, tested on JDK 18 and 11.
This project uses Maven to manage libraries and compile the code. All library dependencies are specified in the pom.xml file.
To compile, navigate to the root directory of the project and run:
mvn package
Successful comilation will produce a jar file in /target/ from which classes that implement a main function can be executed, e.g.
java -cp target/wny-1.0.jar wny.Query
The required external libraries, together with the version tested, are:
- Javatuples 1.2
- z3 4.8.17
- gurobi 9.5.2
- weka 3.8.6
Note that the maven repository does not provide Z3 and Gurobi, therefore the corresponding jar file needs to be obtained and provided locally. Also, using Gurobi requires a license. (Free options are generally available, but this may change over time.)
Experiment.java runs all experiments for the VLDB 2023 paper.
The data folder is used for the synthetic data generation.
The entities folder contains all entity classes including Relation, Tuple, Box and Constraint.
The query folder provides the class for a why-not-yet query.
The solver folder contains the Gurobi and Z3 solvers.
The util folder is used for parsing a database.
All data used in our experiments are here, including the NBA stats data and the synthetic data representing 3 different data distributions.
Some code for the database parser and the tuple and relation data structure are originally from the any-k repository by Nikos Tziavelis. Thank you Nikos for sharing your code!
The NBA data were obtained on October 2, 2022, from the great baseketball stats and history website Basketball Reference. We appreciate and like the website a lot for its detailed stats.
Zixuan Chen (chen.zixu@northeastern.edu)