Shared tasks in natural language processing
Course organizers: Mark Fishel
Code: MTAT.06.046, MTAT.06.047
Load: 6 credit points
Meetings: some Tuesdays at 12:15 in Liivi 2 - 511
Bring it on! A shared task is a competition for automatic natural language processing systems. Typically it involves machine learning, so the participants get the same training and dev sets and use them, as well as a set of rules, hacks and etc. to develop a solution to the task at hand. In the end all the participating systems are evaluated on the same test set.
As part of this Spring's seminar you will pick one of the shared tasks (below) and work on a submission to it.
Why do it? Because the point is in pushing science forward! Solving the unsolved!! Answering the unanswered!!! Doing the undone!!!!11!1ONEONE
More specifically: in case of a successful participation you will have a research paper with your name on it + 6 credit points (3 for MTAT.06.046 and 3 more for MTAT.06.047).
Shared tasks this year:
- WMT:
- neural machine translation; this includes Estonian<->English!! :-)
- multimodal translation
- quality estimation
- translation metrics
- CoNLL
Some time around May the deadlines for the final submissions will strike. In between then and now we must develop, test, fix, test, etc. ad nauseam the submission systems.
Join the dark side!