This Action aims at improving the pre-existing neural machine translation toolkit “Marian” to address the needs of CEF eTranslation and to broaden its user base.
This collaborative language tools Action will provide user access to text flagging by adding support for factors to support forced translation. It will also support the automated adaptation facility by adding support for on-the-fly domain adaptation using large translation memories with a well-defined API.
Furthermore, Marian’s documentation will be improved for all levels of users, resolving requirements to submit to ELRC-SHARE and increasing the user base.
To ensure the dissemination of the results, a stakeholder board and a hackathon will be organised. All the results of the activities will be disseminated through open-source software deliverables and engagement with eTranslation.
Milestones
01/11/2020 Factors Code Integrated
01/03/2021 Interim Documentation: overview of the Internal architecture, documented library API for so users can create graphs
01/05/2021 Forced Translation Code Integrated
01/08/2021 GPU Efficiency Code Integrated
28/02/2022 On-the-fly Domain Adaptation Code Integrated
- MT Half-Marathon starts today!Last chance to join this three-day virtual gathering for Machine Translation researchers, developers, students and …
- WMT2021 Efficiency Shared TaskUser-Focused Marian will take part in WMT2021 Efficiency Shared Task. Check out the task description …
- User-Focused Marian hackathonSign up for User-Focused Marian hackathon here
- How-to on forced translation with MarianCode and documentation on how to do forced translation with Marian is now available! This …
- Marian has new user documentation!Marian has new user documentation! This new documentation covers the expression graph and its operators, …