Consortium

 

 

 

dfkiGerman Research Center for Artificial Intelligence (DFKI)
DFKI is coordinating the QT21 research and innovation action. DFKI has expertise in enriching HPB-SMT with linguistic information, and will be closely involved in WP1, in particular in the systematic search for optimal granularity of HPB-SMT (and other syntax and semantic based statistical translation formalisms). DFKI has led the development of MQM error typology and metrics and will contribute to WP3. DFKI will also lead WP5 based on extensive expertise gained in META-NET.
RWTH_LogoRheinisch-Westfälische Technische Hochschule Aachen (RWTH)
RWTH has long-standing expertise and interest in syntax-enhanced SMT (the JANE HPB-SMT system), the use of neural nets in language technologies and the acquisition of MT resources from mono-lingual data only, and will make key contributions to research in WP1 and WP2.
UvA
University of Amsterdam
The Computational Linguistics Lab at University of Amsterdam has extensive experience relevant for QT21, particularly in hierarchical factorization for reordering in SMT (WP1), syntax-driven SMT (WP1), learning (latent) syntactic labeling for hierarchical SMT (WP1 and WP2), and statistical models for joint morphological and syntactic parsing for morphologically-rich languages (WP2).
DCUDublin City University (DCU)
Because of their long-standing interest in syntax-enhanced MT, DCU will be closely involved in WP1. DCU has been a pioneer in early research on syntactically and semantically enhanced automatic evaluation, and has a key involvement in WP3.
university_edinburghThe University of Edinburgh (UEDIN)
The University of Edinburgh has long standing interest and expertise in syntactically enhanced SMT and morphologically rich languages and will make key contributions to WP1 and WP2. For many years, UEDIN have been a key part of the WMT organisation and will therefore lead the shared task WP4.
kit_logoKarlsruhe Institute of Technology (KIT)
KIT has long standing research expertise and interest in the application of neural nets in natural language processing and will bring to bear that in leading WP2. KIT also explored new ideas on using source side author driven guidance in WP3.
CNRS_fr_quadri.svgFrench National Agency for Scientific Research (CNRS)
LIMSI, a CNRS Laboratory, provides key expertise for Romance languages
and neural networks and will bring this to bear to explore new models
for under-resourced and morphologically complex languages (WP2), as well as for translation quality improvement (WP4).
220px-Carolinum_Logo.svgCharles University in Prague (CUNI)
CUNI is an early pioneer in the use of deep linguistic models (syntax and semantics) in rule-based and statistical approaches to MT. CUNI is also an expert in morphologically rich languages (Slavic, Czech) and will bring this to bear leading WP1 and making a strong contribution to WP2.
Fondazione-Bruno-Kessler_mediumFondazione Bruno Kessler (FBK)
FBK has developed strong expertise and interest in the new research area of exploiting knowledge gained from post-editing operations in the MateCat project and will bring this to bear leading WP3.
UniSheffieldThe University of Sheffield (USFD)
USFD has strong expertise in quality estimation and will bring this to bear in WP3. USFD is also developing models for optimally selecting training instances for MT and will apply this in resource scarce settings in WP2.
taus_bvTAUS BV
TAUS has developed the Dynamic Quality Framework (DQF) and will assume a key role in research on error typologies and evaluation metrics in WP3. Due to its extensive translation industry PR and outreach operation, TAUS will also be involved in a key role in WP5.
text-form-logotext&form GmbH
text&form has been a key partner in developing the MQM error typology and evaluation metrics in the QTLaunchPad project. It will make this expertise bear on WP3. text&form is also a key contributor of professional human translators for morphologically complex languages for annotation, postediting and evaluation.
Tilde_logoTilde Sia
Tilde provides key LSP expertise in developing MT solutions for challenging languages and resource scenarios (Latvian and Lithuanian, and other Baltic languages) and will concentrate on WP3 and WP2. They are also a major source for human professional translation expertise for annotation, post-editing and evaluation (WP3 and WP4).
hongKong_standard_logoHong Kong University of Science and Technology (HKUST)
Since 1992 the Human Language Technology Center (HLTC) at HKUST has spearheaded international research in SMT with the first tree-based syntactic SMT models, and then semantic SMT models, and more recently, semantic machine translation evaluation. The recent stream of work on semantic frame based MT evaluation is state-of-the-art under both human and automatic approaches creating directly relevant foundations for WP3 and WP4.