If you want to participate in the course, plase join the Moodle page of the course.
Reproducibility of experiments is one of the essential components of the modern science. Since most work in Computational Linguistics (CL) and Natural Language Processing (NLP) heavily rely on experiments (typically with Machine Learning methods), reproducibility is also an important concern for computational linguists. Reproducibility is not only relevant to (scientific) research. It is also important for most, if not all, “natural language engineering” tasks: it is crucial for any practical NLP method to show as good performances in real-world usage as the performance observed during development.
In the first part of the course, we will review some of the basic concepts behind reproducibility in science, focusing on reproducibility of methods used in machine learning and NLP. The second part of the course will take a more practical approach. The participants are required to choose a published article in CL or related fields, and reproduce the results presented in the article.
The course can be taken either as a 6CP HS (participation, project and project presentation) or as a 9CP course with an additional of term paper describing the project.