01
Provider fragmentation
Translation engines used different APIs, formats, authentication methods and operational environments.
Secure European language infrastructure
iADAATPA and MT Hub: one secure access point for multiple translation engines
Led by Pangeanic, iADAATPA developed an open platform that allowed European public administrations to access, compare and route translation requests across several machine translation providers through a common secure infrastructure.
2016 EU IA 0132 Connecting Europe Facility action identifier
2017 to 2019 Development, testing and public platform launch
6 core partners Technology companies, systems integration and academic expertise
Open source Platform code published under the GNU GPL v3 licence
Pangeanic led Architecture, security oversight and consortium leadership
The public sector challenge
Public administrations needed access to translation technology without becoming dependent on one provider
Machine translation services differed by language, subject area, deployment model and quality. A provider that performed well for legal Spanish might not offer the best engine for Lithuanian parliamentary content or tourism data.
Integrating every provider separately created duplicated development, incompatible APIs and long procurement cycles. Switching between systems could require new connectors, security assessments and workflow changes.
iADAATPA addressed this fragmentation by creating a common platform able to receive translation requests, identify their language and domain, and route them to the most appropriate available engine.
02
Uneven model performance
Quality varied according to language direction, subject domain, training data and engine specialization.
03
Sensitive public data
Institutional documents required secure transport, controlled access and infrastructure suitable for public sector use.
04
Integration overhead
Every additional provider could introduce new connectors, maintenance requirements and workflow complexity.
The MT Hub concept
Separate access to language technology from dependence on a single engine
MT Hub created a shared access and orchestration layer between public administration workflows, European infrastructure and multiple machine translation providers.
01
Receive the request
Documents, text and web content enter the platform through a common interface, connector or API.
02
Detect the language
Automatic language identification determines the source language before translation begins.
03
Identify the domain
Domain detection classifies content so that specialized engines can be selected where available.
04
Route intelligently
Translation requests are sent to the most suitable provider according to language, domain and service configuration.
05
Exchange securely
AS4 Domibus and European eDelivery components support reliable transfer of documents and service messages.
06
Return the result
Translated content is delivered through the same platform, reducing integration work for the requesting administration.
Model and provider orchestration
One interface, several engines and a routing layer between them
The platform treated machine translation engines as interchangeable services that could be selected according to operational requirements.
Document or translation request
Content arrives through an API, CAT tool, content management system, website or institutional workflow.
→
Detection, policy and routing
Language, domain, provider availability and service configuration determine the translation path.
→
Infrastructure capabilities
A complete operational layer around machine translation
iADAATPA combined translation technology with the integration, routing, security and workflow components required for institutional use.
ROUTE
Automatic request routing
Select engines according to source language, target language, domain, availability and configured service rules.
DOMAIN
Automatic domain detection
Classify content before translation to improve selection of specialized models and terminology.
LANG
Language identification
Detect the source language automatically when the requesting workflow does not provide reliable metadata.
AS4
Secure eDelivery
Use AS4 Domibus components for reliable and secure exchange aligned with European digital infrastructure.
API
Synchronous and asynchronous APIs
Support immediate text translation and longer document processing through different request patterns.
WEB
Private website translation
Browse and translate web content through controlled platform services without exposing user activity unnecessarily.
CAT
CAT tool connectors
Connect translation environments including Trados, MateCat and OmegaT to the same shared infrastructure.
OPEN
Open source platform
Publish the routing platform openly so public bodies and technology providers could inspect, deploy and extend it.
Pangeanic in iADAATPA
Consortium leadership, platform security and multilingual infrastructure design
Pangeanic led the consortium and oversaw the architecture required to connect several commercial technologies with European public administration services.
LEAD
Consortium leadership
Pangeanic coordinated commercial providers, systems integrators, academic participants and institutional users.
ARCH
Architecture oversight
The company supervised development of a provider neutral platform capable of handling multiple engines and workflows.
SEC
Security requirements
Pangeanic was responsible for security features and alignment with public sector document exchange requirements.
MT
Neural translation services
Pangeanic contributed specialized machine translation capacity alongside the other technology providers.
TEST
Institutional validation
Real public sector and media use cases tested routing, domain detection and secure translation workflows.
OPEN
Open infrastructure
The resulting platform was released as open source to support reuse beyond the original project.
Institutional pilots
Tested across government, tourism, parliamentary, educational and media environments
The platform was evaluated through operational settings with different language pairs, content domains and integration requirements.
Spain
Generalitat Valenciana
Translation workflows connected to the regional government Open Data and Transparency portal.
Spain
SEGITTUR
Multilingual digital services supporting tourism technology and public information.
Lithuania
Lithuanian Parliament
Translation of parliamentary proceedings and institutional content involving a lower resource European language.
Ireland
Dublin City University
Multilingual website and academic content used to test platform access and translation delivery.
Germany
Deutsche Welle
International media content used to validate multilingual routing and translation at practical scale.
Integration without lock in
Connect existing tools once and access several translation services
The value of MT Hub extended beyond its translation engines. It created a common integration layer so that institutional users could keep their preferred tools while accessing different providers.
Connectors and APIs reduced the need to rebuild workflows every time an engine changed. This enabled public bodies to compare providers, introduce specialized models and maintain continuity at the application layer.
The same principle applies to modern enterprise AI. Models will change more rapidly than the systems and processes that depend on them. A controlled orchestration layer protects the organisation from unnecessary provider dependence.
REST API and SDK Application and service integration
SDL Trados Professional translation workflows
MateCat Browser based CAT environment
OmegaT Open source translation environment
Drupal Multilingual content management
OpenCMS Institutional website publishing
Browser plugins Direct multilingual web access
AS4 Domibus Secure European document exchange
Project results
A working multilingual service layer for European public administrations
iADAATPA moved beyond a translation engine demonstration and delivered the infrastructure required to manage several engines as a common public service.
1
Common access point
Public administrations could reach several translation services through one platform and integration layer.
Multi
Provider architecture
European Commission and commercial engines could operate behind a shared request and routing mechanism.
AS4
Secure document exchange
European eDelivery components supported reliable communication between institutional systems.
GPL
Open source release
The platform was made publicly available under the GNU General Public License version 3.
European consortium
Translation providers, systems integration and academic language technology
The consortium brought together competing machine translation providers inside a shared architecture, supported by systems integration and academic expertise.
Consortium lead
Pangeanic
Consortium leadership, security oversight, architecture supervision and neural translation services.
Technical implementationKantanMT
Core platform coding, hosting and integration of commercial machine translation services.
Language technologyTilde
European language technology, specialized engines and involvement of the Lithuanian Parliament.
Language technologyPrompsit
Machine translation services, language processing expertise and provider integration.
Systems integrationEveris
Systems integration, testing, project management support and identification of deployment scenarios. Everis is now part of NTT DATA.
Academic partnerDublin City University
Academic language technology expertise and institutional website validation.
MT Hub in 2026
An early blueprint for sovereign model orchestration
Modern organisations increasingly use several models rather than placing every task inside a single general system. Different languages, domains, security levels and workflows require different capabilities.
iADAATPA anticipated this environment. It placed a neutral control layer between users and models, allowing engines to be selected according to policy, performance and context.
The principle now extends beyond machine translation. Enterprises require governed gateways that can route tasks across small language models, private models, retrieval systems and external services while retaining control over data flows.
ROUTE
Model routing
Select the most appropriate model according to task, language, domain, quality and deployment policy.
OPEN
Reduced lock in
Keep applications separate from individual providers through common interfaces and orchestration.
SEC
Controlled data flows
Apply security, routing and infrastructure rules before documents are sent to a model.
SLM
Specialized models
Use smaller or domain adapted systems where they offer better operational performance.
From public infrastructure to enterprise AI
What MT Hub demonstrates for organisations operating multiple AI systems
The project provides a practical precedent for controlled access to several models through a shared enterprise layer.
01
For public administrations
Connect multilingual services to secure institutional workflows without rebuilding every application around one vendor.
02
For regulated enterprises
Route sensitive content according to location, provider, model, security and data retention policies.
03
For enterprise AI teams
Manage specialist and general models behind consistent APIs, evaluation rules and operational controls.
04
For sovereign AI programmes
Maintain control over model selection, data movement, infrastructure and service continuity.
Public evidence
European Commission documentation, open source code and project records
These references document the action identifier, architecture, security layer, pilots and public release of MT Hub.
European Commission
eTranslation and eDelivery used to build a machine translation router
European Commission account of the platform architecture, eTranslation integration, Domibus security and pilot environments.
Read the Commission article →Open source code
iADAATPA repositories on GitHub
Public repositories containing components released by the project under an open source licence.
Explore the repositories →Project completion
iADAATPA becomes MT Hub
Pangeanic project account covering the consortium, platform capabilities, launch and open source release.
Read the project record →Platform concept
Open source multi provider MT platform
Early explanation of the common hub model, connectors, APIs and provider independence.
Read the platform article →European Language Data Space
ELRC workshop report for Spain
European report describing iADAATPA as the project that produced MT Hub, an open source routing platform.
View the report →Pangeanic project portfolio
European AI and language technology projects
MT Hub within Pangeanic’s wider work in multilingual data, translation infrastructure, privacy and sovereign AI.
Explore all projects →Continue exploring
From translation orchestration to multilingual models, privacy and reusable language data
iADAATPA established an infrastructure layer that was extended through later European projects addressing models, data assets and sensitive information.
European project portfolio
Explore Pangeanic’s European AI and language technology projects
Review projects covering multilingual datasets, translation, privacy, cultural heritage and sovereign AI infrastructure.
View all projects → Neural model infrastructureNTEU
Discover how Pangeanic coordinated 506 direct neural translation directions between the official EU languages other than English.
Explore NTEU → Privacy infrastructureMAPA
Explore multilingual anonymization models and annotated datasets for sensitive administrative, legal and medical documents.
Explore MAPA → Reusable language dataNEC TM
Discover the European infrastructure created to recover, share and reuse translation memories generated by public administrations.
Explore NEC TM →Sovereign language and model orchestration
Operate multilingual AI without placing every workflow inside one provider
Pangeanic supports enterprises and public institutions with private translation infrastructure, model routing, multilingual data, controlled APIs, evaluation and sovereign deployment.