„eTranslation TermBank“ – naujas tarptautinis projektas Europos Komisijos automatinio vertimo kokybei pagerinti

eTranslation TermBank – a new international project to improve the European Commission’s machine translation quality EGLĖ PESLIAKAITĖ Institute of the Lithuanian Language On 1 September 2017, a new international project, “eTranslation TermBank: Collecting and Providing Domain- and Language-Specific Terminology Resources for the Public Digital Services Sector,” was launched. The project’s lead coordinator is Tilde (Latvia). Representatives of eight institutions from Lithuania (Institute of the Lithuanian Language), Denmark (University of Copenhagen), Estonia (Institute of the Estonian Language), Slovenia (Jožef Stefan Institute), Austria (International Terminology Network (TermNet)), Iceland (Árni Magnússon Institute for Icelandic Studies), and Sweden (Institute for Language and Folklore, Swedish Centre for Terminology) are also participating in the project. The aim of the project is to improve the quality and coverage of the European Commission’s machine translation. MACHINE TRANSLATION – REDUCING LANGUAGE BARRIERS IN EUROPE Linguistic diversity is a great asset of the European Union, but in today’s digital world multilingualism often creates obstacles: language barriers make not only communication more difficult, but also cooperation between the citizens and countries of the European Union. Machine translation (English eTranslation) is one of the services funded under the European Infrastructure and Networks Facility (EITP) (English Connecting Europe Facility (CEF)) programme, which aims to connect missing links between Europe’s energy and transport sectors and digital resources. The translation service has been available since 2013 to EU institutions, Member State administrations, and several European Commission information systems and online services. This translation is based on statistical machine translation (MT) technology, which learns to translate from available translations produced by people (for more information, see European Language Resource Co- Terminologija | 2017 | 24 205 ordination. European language resource coordination, available online: http://www.lr-coordination.eu). With machine translation, users can instantly translate words, sentences, documents, and websites from one language into another. Most importantly, the machine translation system can also be connected to any public administration digital service infrastructure required for European cooperation (English Digital Service Infrastructures (DSIs)), such as e-health services, an open data portal, an online dispute resolution portal, and so on, making that infrastructure multilingual. However, to achieve effective results that are useful to European public service administrations, public service users, and society—namely, high-quality translation and the ability to use services in one’s mother tongue—as many suitable language resources as possible need to be collected. It was to pursue this goal that the partners in the eTranslation TermBank project joined forces. PROJECT OBJECTIVES AND PLANNED RESULTS The project eTranslation TermBank is overseen by the EITP-supported European Language Resource Coordination Network and its members of the Language Resource Board (national representatives in each European Union country), who provide information about the terminology resources available in each country. Over the course of the eighteen-month project, terminology resources in specific fields are to be collected in all European Union languages (as well as Norwegian and Icelandic) and made available to the public digital services sector. The European Commission is currently seeking to strengthen public digital services in several fields that require machine translation, namely health (e. health), business law (company registration and deregistration (e. law)) and consumer protection (online dispute resolution). The project will be implemented in three stages: 1. Identification of terminology resources. 2. Assessment of the identified resources. 3. Provision of suitable terminology resources to the machine translation platform. Since it is important for machine translation to be useful throughout Europe and to translate from every European Union language into every other, the project’s nine partners have divided up the European languages whose terminology resources will be identified, collected, assessed and provided for machine translation. During the project, the Lithuanian representatives will work with terminology resources in Lithuanian, Polish and Italian. 206 Eglė Pesliakaitė eTranslation TermBank – naujas tarptautinis projektas Europos Komisijos automatinio vertimo kokybei pagerinti In pursuing the project’s objectives, the partners will cooperate with national institutions across Europe that work with terminology and have the necessary resources. The project plans to collect 150 new terminology resources in all the languages of the European Union, as well as Norwegian and Icelandic, that have not yet been registered in European Commission databases or incorporated into the machine translation system. At the initiative of the project coordinators, the machine translation system will be enhanced with these new terminology resources, thereby improving the quality of machine translation and expanding the opportunity to make full use of the public services provided by the European Commission. Catalogues will also be compiled of national institutions working with terminology and of terminology resources covering specific fields, and a methodology will be provided for preparing and applying terminology resources for machine translation. Thus, the project will also be extremely valuable in expanding and strengthening cooperation among European terminologists and institutions working with terminology resources. eTranslation Termbank – A New International Project to Improve the Quality of Automated Translations for the European Commission 1 September 2017 saw the start of a new international project called eTranslation TermBank: collection and provision of terminological resources in sector-specific domains and languages of interest to sector-specific digital public services. The key coordinator for this project is Tilde (Latvia); plus, there are eight other partners involved in the project: the Institute of the Lithuanian Language (Lithuania), the University of Copenhagen (Denmark), the Institute of the Estonian Language (Estonia), the Jožef Stefan Institute (Slovenia), the International Network of Terminology, TermNet (Austria), the Árni Magnússon Institute for Icelandic Studies (Iceland), and the Institute of Language and Folklore and the Swedish Centre for Terminology (Sweden). The project aims to improve the quality and coverage of automated translations for the European Commission. Terminologija | 2017 | 24 207 ETRANSLATION TO BRING DOWN LANGUAGE BARRIERS IN EUROPE Linguistic diversity is a key value of the European Union, however in today’s digital world multilingualism often creates obstacles and language barriers that make it more difficult for the citizens and Member States of the European Union to communicate and cooperate. Automated translation, or eTranslation, is a service financed under the Connecting Europe Facility (CEF) that aims to provide the missing links between energy, transport domains, and digital resources in Europe. The translation service was launched in 2013 and is available to EU institutions, Member State administrations, and several European Commission information systems and Internet services. This type of translation is based on the machine translation (MT) technology, which learns to translate from the available pool of man-made translations (for details, see European Language Resource Coordination, online access at: http://www.lr-coordination.eu). Automated translation allows users to translate words, whole sentences, documents, and websites from one language to another. The most important thing is that the eTranslation system can also be customised and connected to render multilingual any public administration digital service infrastructure (DSI) needed to enable cooperation in Europe, such as e-health services, the open data portal, the Online Dispute Resolution (ODR) portal, and so on. Still, to achieve outcomes that are both effective and beneficial to European public service administration institutions and public service users, which outcomes would constitute high-quality translations and a possibility to use the services in one’s mother tongue, we have to accumulate as many adequate language resources as possible. It is to attain that purpose that the partners of the eTranslation TermBank project joined forces. PROJECT OBJECTIVES AND DELIVERABLES The eTranslation TermBank project is supervised by the European Language Resource Coordination network and its members from the Language Resources Council (national representatives in every Member State) supporting the collection of data in each country. The eighteen-month project is geared towards collection and provision of sector-specific terminological resources for all European Union (as well as Norwegian and Icelandic) languages to sector-specific digital public services. Currently, the European Commission seeks to strengthen the digital public services in several domains that require automated translation, such as health 208 Eglė Pesliakaitė │ eTranslation TermBank – naujas tarptautinis projektas Europos Komisijos automatinio vertimo kokybei pagerinti (e-Health), business legislation (company registration and de-registration; e-Justice), and consumer protection (ODR – Online Dispute Resolution)). The project will be implemented in three stages: 1. Identification of the required terminological resources. 2. Assessment of the resources identified. 3. Provision of the appropriate terminological resources to the eTranslation platform. Since it is critical for automated translation to benefit the entire Europe and be able to provide translations from and to every European Union language, the nine partners in the project have divided among themselves the European languages whose resources they will be identifying, compiling, and providing for automatic translation. The Lithuanian party will work with Lithuanian, Polish, and Italian language terminological resources in this project. To achieve the objectives of the project, the partners will cooperate with pan-European national bodies that work with terminology and possess the required resources. A total of 150 new terminological resources for all European Union languages, as well as Norwegian and Icelandic, that have never been registered in the European Commission’s databases or included in the automated translation system are to be collected in the course of implementation of the project. At the initiative of the project coordinators, the eTranslation system will be enhanced with the new terminological resources, which will allow improving the quality of automated translation and secure the opportunity for the users to fully enjoy the European Commission’s public services. In the process of delivering the project’s objectives, a catalogue of national institutions working with terminology and a catalogue of sector-specific terminological resources will be made and a methodology of preparation and application of terminological resources for automated translation drafted, which will greatly increase the project’s value for expanding and reinforcing the fine-tuned cooperation among European terminologists and institutions that work with terminological resources. The article was translated into English by Jūras Barauskas Received 2017-10-30 Eglė Pesliakaitė Institute of the Lithuanian Language 5 P. Vileišio St., LT-10308 Vilnius Email: [email protected] Terminologija | 2017 | 24 209