Detailed Timeline for Organizers
Color coding of deadline addressees: participants – “ordinary” data providers – specific data processing – organizing committee.
December 11 (Sunday): Announcement of the shared task and set up of the shared task website. Registration for the Shared Task open (by e-mail unless we manage to set up a web registration form).
Beginning of January: Translations done by DFKI should be available.
January 10 (Tuesday): Deadline for suggesting additional data by participants (registration necessary).
January 15 (Sunday): “Commitment deadline” for data contributors – tell the organizers that you intend to have your data included.
February 15 (Wednesday): The raw data from CommonCrawl, prepared by Turku, must now be available for Milan to start preparing UDPipe (no embeddings needed at this moment) and to train the detokenizer for languages lacking SpaceAfter=No.
February 15 (Wednesday): Data freeze for UD v2 data that is going to be included in the shared task. UDPipe will now be re-trained on the new data and applied to them, cross-validation style, so we can release predicted morphology too.
February 20 (Monday): Trial data publicly available.
February 28 (Tuesday): Task Registration deadline. Participants have to register to setup their evaluation space and other data, and get access to task data later.
March 1 (Wednesday): Release of training + development data. That includes versions of training + development data with annotation predicted by UDPipe, and also the raw data with pre-computed word embeddings (Turku) and preprocessing by UDPipe (Milan).
March 31 (Friday): Google annotation of parallel data from DFKI available.
April 16 (Sunday): Parallel test data converted to v2 by Dan, language owners should now check them.
April 23 (Sunday): A preliminary version of the surprise language test data, with manual POS tags, should be available for Milan to start training hyper-parameters.
April 30 (Sunday): Data-freeze for the parallel test data and for the surprise language test data. All languages that did not take the mainstream path DFKI – Google – Dan – final check, must now be available in UDv2 too. Milan has now 1 week to process the data by UDPipe.
May 1 (Monday): Descriptions of surprise languages and sample data released. If we want to provide raw data for the surprise languages, it must be ready by now too.
May 8 – 12 (Monday – Friday): Test phase. We could make the test data available some time between Friday evening and Monday morning (preferably before European midnight Sunday-Monday). The deadline for the system outputs could be Friday 23:59 Samoa Standard Time (UTC-11, i.e. Saturday 12:59 in Central Europe).
May 15 (Monday): Results announced (someone has to work over the weekend).
May 26 (Friday): Submission of papers (we have negotiated a deadline extension because the other ACL workshops will probably have their deadline on Friday May 19).
June 2 (Friday): Reviews due.
June 9 (Friday): Final papers due. (While the other workshops probably will have the camera-ready deadline two weeks earlier, Friday May 26.)
August 3 – 4 (Thursday – Friday): CoNLL conference, Vancouver, Canada