MT Community, LDC training data for OpenMT12 are now available; please see the LDC License Agreement linked on the OpenMT12 website, http://www.nist.gov/itl/iad/mig/openmt12.cfm. The license agreement should be filled out and sent directly to the LDC when you register for OpenMT12. The resources in the license agreement include Arabic, Chinese, Korean, and English. Note that NIST can make available a limited amount of Dari and Farsi data separately for OpenMT12, as outlined in the evaluation plan.
Please note: If you register for OpenMT12 and do not wish to receive training data under the OpenMT12 agreement with the LDC, you still must fill out the agreement and minimally check off the OpenMT12 test sets for the language pairs you register for, at the bottom of the list of resources. Those that have already registered for OpenMT12 should also return the license agreement to the LDC at this time. NIST cannot send out the test sets at evaluation time without the license agreement for the test data in place. Best, Kay. From: Peterson, Kay Sent: Friday, October 28, 2011 5:22 PM To: mt_l...@nist.gov; mt-list@eamt.org Subject: Announcing NIST OpenMT12 MT community, I am pleased to announce NIST's next Open Machine Translation evaluation, OpenMT12. The OpenMT12 website is now available at http://www.nist.gov/itl/iad/mig/openmt12.cfm, with links to the evaluation plan and necessary forms. The evaluation plan will soon be updated to specify the submission requirements formally; please check the website regularly for updates. Registration for OpenMT12 is now open, and we highly encourage early registration. Highlights of OpenMT12: - Evaluation on prior years' Arabic-to-English and Chinese-to-English Progress test data, with the data to be made available to participants after OpenMT12, - Evaluation on a parallel data set based on prior years' Progress tests for Arabic, Chinese, Dari, Farsi, and Korean to English, in two source data styles, - Evaluation on new domain data for Chinese-to-English, - Support for both a single system and a system combination track (provided sufficient interest by the participants), - Evaluation by automatic metrics and coordination of volunteer human assessments using a new tool to rank alternative translations. Schedule: - (See Table 1 of eval plan): Training data off-limits periods. - October 28 2011: Evaluation plan available - October 28 2011 - February 3 2012: Registration period (early registration highly encouraged) - November 2011 (tentative): Training data available from LDC - January 16 - February 24 2012: Dry run period (early submission highly encouraged) - April 2 - 6 2012: Main evaluation period; output due at NIST April 6 - April 16 - 20 2012: Post-eval system combination evaluation period; output due at NIST April 20 - April 20 2012: Preliminary release of main evaluation results to participants - April 27 2012: System descriptions due - April 30 - June 1 2012: Human assessment period - June 27 - 28 2012: Workshop in the Washington DC area, co-located with the NIST OpenHaRT12 evaluation workshop - August 31 2012: Official public release of results Please direct any questions to mt_...@nist.gov<mailto:mt_...@nist.gov>. We are hoping for a lively 2012 evaluation! Best, Kay Peterson NIST ITL IAD Multimodal Information Group http://www.nist.gov/itl/iad/mig Phone (+1) 301-975-2960
_______________________________________________ Mt-list mailing list
