Managing Public Sector Data for Multilingual Applications Development

Managing Public Sector Data for Multilingual Applications Development Stelios Piperidis, Penny Labropoulou, Miltos Deligiannis, Maria Giagkou Athena RC/Institute for Language and Speech Processing Epidavrou & Artemidos, Maroussi, Athens, Greece spip@ilsp.gr, penny@ilsp.gr, mdel@ilsp.gr, mgiagkou@ilsp.gr Abstract The current paper outlines the ELRC-SHARE repository, an infrastructure designed and developed in the framework of the European Language Resource Coordination action with the objective to host, document, manage and appropriately distribute language resources pertinent to machine translation, and specifically tailored to the needs of the eTranslation service of the European Commission. Due to the scope of the eTranslation service which seeks to facilitate multilingual communication across public administrations in 30 European countries and to enable Europe-wide multilingual digital services, ELRC-SHARE demonstrates a number of characteristics in terms of its technical and functional parameters, as well as in terms of its data management and documentation layers. The paper elaborates on the repository technical characteristics, the underlying metadata schema, the different ways in which data and metadata can be provided, the user roles and their respective permissions on data management, and, finally, the extensions currently being implemented. Keywords: public sector language resources, repository, documentation, machine translation 1. Introduction Machine Translation (MT) technology can be applied to any language pair and adapted to specific domains and text types provided that sufficient amount of Language Re- sources (LRs) can be made available for training, adapt- ing and evaluating the respective translation engines. The eTranslation digital service infrastructure1 developed by the European Commission (EC), for example, needs to be adapted to meet the quality requirements of other Digital Service Infrastructures (DSI) like eJustice or Online Dis- pute Resolution, besides being extended to new language pairs. To meet the objectives of eTranslation adaptation, extension and overall improvement, the EC has launched the European Language Resources Coordination (ELRC) action (L¨osch et al., 2018) with the mandate to collect and manage the appropriate LRs for all languages of the EU plus Norwegian and Icelandic, and for a number of domains that are relevant to other DSIs that the EC is building under the umbrella of the Connecting Europe Facility (CEF) pro- gramme2. Properly collecting, documenting and managing such data and making them available for developing MT en- gines calls for appropriate LRs management supported by a dedicated repository with the appropriate functionalities catering for the whole LRs lifecycle. In this paper we focus on the repository infrastructure de- veloped for this initiative, i.e. the ELRC-SHARE repos- itory. Section 2 briefly introduces the ELRC action and provides an overview of the ELRC-SHARE technical char- acteristics, section 3 describes the underlying metadata schema used for LRs documentation, section 4 explains the different ways in which data, i.e. LRs and their metadata can be provided and managed, section 5 elaborates on the user management layer, while section 6 describes some of the extensions currently being built. 1https://ec.europa.eu/cefdigital/wiki/display/CEFDIGITAL/ eTranslation 2https://ec.europa.eu/digital-single-market/en/news /connecting-europe-facility-cef-digital-service-infrastructures 2. ELRC and ELRC-SHARE 2.1. About the ELRC action Within this framework, the ELRC Network aims to raise awareness, promote uptake and foster the acquisition, iden- …

Content truncated to 500 words