Projects

We are developing statistical machine translation systems for a wide range of language pairs, and explore many research questions related to translation.

We are currently funded by the following projects:

Current Projects under EU H2020 ICT 17 Framework

QT21, 2015-2017

QT21 is an ambitious research project which addresses the challenge of producing significantly improved machine translation models for language pairs that include “difficult” languages, focusing on morphologically rich and syntactically divergent languages and languages that are under-resourced.

MMT (Modern Machine Translation) , 2015-2017

The goal of MMT is to deliver a large-scale commercial online translation infrastructure based on a new open-source distributed MT architecture.

HimL (Health in my Language), 2015-2017

HimL aims to increase timely access to public health information by focussing on high accuracy translation. It will develop domain adaptation techniques, semantically aware models and evaluation metrics, and tackle morphologically rich target languages.

TraMOOC (Translation of Massive Open Online Courses) , 2015-2017

The goal of TraMOOC is to leverage the educational advantage of MOOCs across Europe, developing high-quality translation of all types of text genre (e.g. assignments, tests, presentations, lecture subtitles, blog text) from English into eleven European and BRIC languages.

Cracker, 2015-2017

Cracker is a co-ordinating action responsible for running evaluation campaigns and outreach and development activities for the translation community.

SUMMA

Previous Projects

Accept, 2012-2014

This project is concerned with the translation of User-generated Content (UGC).

Casmacat, 2011-2014

The CASMACAT project will build the next generation translator's workbench to improve productivity, quality, and work practices in the translation industry.

EU-Bridge, 2012-2014

EU-BRIDGE aims at developing automatic transcription and translation technology that will permit the development of innovative multimedia captioning and translation services of audiovisual documents between European and non-European languages.

MateCat, 2011-2014

The goal of MateCat is to create new CAT technology that will significantly enhance the productivity and user experience of professional translators.

MLT4MLV , 2011-2013

This project addresses the translation of closely related language varieties, specifically Standard Austrian German and Viennese Dialect.

MosesCore, 2012-2015

In this Coordination Action, we aim to suppport the development and use of open-source translation tools through machine translation marathons, shared tasks and workshops, and industrial outreach.

EuroMatrixPlus, 2009-2012

EuroMatrixPlus builds on the success of the EuroMatrix project. It is funded under Framework Programme 7 by the European Commission. The project's goal is to advance machine translation technology and bringing it to the user.

LetsMT!, 2010-2013

LetsMT! is funded Framework Programme 7 by the European Commission. It's goal is to build a web service that allows users to upload translated texts and automatically build customized machine translation systems.

AGILE, 2005-2011

We are member of the AGILE consortion, which is a project for machine translation of speech and text from Arabic and Chinese. This is a joint project with BBN, ISI, MIT, and Cambridge University (and other partners working on speech recognition and distillations). It is funded by the US Defense Advanced Research Program Agency (DARPA), under a funding program called GALE.

EuroMatrix, 2006-2009

EuroMatrix is a Framework Programme 6 project funded by the European Commission on text translation between all official languages in the Europen Union. This is a joint project with the University of Saarbrücken, Charles University (Prague), CELCT, Linear B, GlobalWare and MorphoLogic. The technical coordination of this project is with the University of Edinburgh.

Demeter, 2007-2009

Demeter is a three year, EPSRC-funded project looking at large-scale (aka 'millions of features') discriminative training for SMT.

RandLM, 2008

RandLM? is a 6 month project funded by Google, looking at randomised language models.

print
Page last modified on April 25, 2017, at 04:25 PM