This is the course page for the seminar course Challenges in Computational Linguistics at the Department of Linguistics, University of Tübingen.

The filed of computational linguistics has a well-established tradition of “challenges” or “shared tasks” where the participants try to solve a current problem in the field using a common data set and a well-defined metric of success. Participation in these tasks is fun and highly educational as it requires the participants to put all their knowledge into practice, as well as learning and applying new methods to the task at hand. The comparison of the participating systems at the end of the shared task is also a valuable learning experience, both for the participating individuals and the whole field.

This course takes its title literally. The students taking the course are required to participate in a shared task in the field, and solve it as best as they can. The requirement of the course include developing a system to solve the problem defined by the shared task, submitting the results and writing a paper describing the system.

Requirements

The course requires good programming skills, a working knowledge of machine learning and NLP, and strong (self) motivation. This typically means a highly motivated Mater’s or advanced Bachelor’s student in computational linguistics or related departments (e.g., computer science, artificial intelligence, cognitive science). If you are unsure, please contact the instructor.

Shared tasks

In principle, any shared task related to computational linguistics is acceptable. Here are a few pointers:

• SemEval competitions include a wide range of shared tasks.
• CoNLL hosts regular high-profile shared tasks.
• Universal Dependencies project organizes multi-lingual dependency parsing competitions. Last two instances were associated with the CoNLL.
• Similarly, SIGMORPHON has regular shared tasks. Again, last two shared tasks were associated with CoNLL.
• Germeval shared tasks are another option, particularly if you’d like to work on German.
• VarDial workshops host regular shared task related to dialects and closely related languages.
• SMM4H is another workshop with regular associated shared tasks.
• CLPSYCH is yet another interesting venue with somewhat regular shared tasks..
• IWCS 2019 has a new shared task on semantic parsing.

The above is only a sample. There are more shared tasks you can find through a search on the Internet. For the purposes of the class, we prefer a shared task where you finalize your work with a system description paper. If all else fails, or if you have a strong preferences, a CL-related Kaggle competition may also be an option (you will still be required to write a system description paper).

Contact

• Instructor: Çağrı Çöltekin <ccoltekin@sfs.uni-tuebingen.de>
  Office hours: Monday 14:00 - 15:00 (Wilhelmstr. 19, room 1.09)