In the ideal world, every CS research paper should be published with the relevant source code and data. This is good for both the original authors (you and your paper are more likely to become popular) and everyone else (easier to reproduce and compare against a previous method).
If you are a member of the latter, look for public domain code and data first.
You can download these without having the authors knowing.
(If you worry about IP tracking, use some VPN tools to hide your trail.)
If you cannot find any similar or related code, email the authors telling them how much you admire their paper and would like to try experiment with their code and/or data.
There is no need to mention anything about your research, or even the fact you are doing a research project.
When you submit your paper, it is usually not possible for the authors to figure out you are even using their code, especially for venues with double blind review processes.
If your paper is accepted for publication, make sure you acknowledge all authors who kindly share the code or data (either publicly or privately with you only), and cite their papers as a good gesture.
If none of the above route works, you might have to reproduce the code.
Do this only as a last resort, as it not only takes time and resources but also poses the question of whether your reproduction is correct.
Leave a Reply