Description du projet
Un nouvel outil convertit les sites web monolingues en d’autres langues
Internet est un système mondial de réseaux informatiques interconnectés. Pourtant, il est essentiellement développé en anglais. Dans l’UE, où le multilinguisme constitue l’un des principes fondateurs, les sites web et les services en ligne destinés aux citoyens ont développé des ressources dans les langues nationales locales, et ne proposent une deuxième langue (généralement l’anglais) qu’en cas de besoin. Cela s’avère insuffisant. Il convient de disposer de nouveaux outils permettant des transformations (et non des traductions) de haute qualité entre les langues. C’est dans ce contexte que le projet EMBEDDIA, financé par l’UE, entend développer la technologie (intégration interlinguistique couplée à des réseaux neuronaux profonds) qui permettra d’utiliser les ressources monolingues existantes dans plusieurs langues sans devoir recourir à d’importantes ressources informatiques.
Objectif
Access to the internet is no longer a luxury---it is a basic component of everyday life and civic engagement, but one in which language continues to be a challenge for fair and equitable access. As Europe becomes more multicultural, and personal and professional mobility between cultures rapidly increases, access to fundamental resources such as local news and government services is limited by the great diversity of the EU's 37 languages. The internet mostly developed in English, and without clear planning for how language issues might form barriers to access and engagement, nor how multilingualism might be supported. In the EU, websites and online services for citizens have developed national local language resources, and often only provide a second language (usually English) when absolutely needed; but the great proliferation of web content, multiple and fast-changing content streams, and an expanding user interest base make this approach untenable. And while advanced natural language research and resources exist for a few dominant languages (English, French, German), many of Europe's smaller language communities---and the news media industry that serves them---lack appropriate tools for multilingual internet development. For the EU to realise a truly equitable, open, multilingual future internet, new tools allowing high quality transformations (not translations) between languages are urgently needed. The EMBEDDIA project seeks to address these challenges by leveraging innovations in the use of cross-lingual embeddings coupled with deep neural networks to allow existing monolingual resources to be used across languages, leveraging their high speed of operation for near real-time applications, without the need for large computational resources. Across three years, the project's six academic and four industry partners will develop novel solutions including for under-represented languages, and test them in real-world news and media production contexts.
Champ scientifique
Programme(s)
Régime de financement
RIA - Research and Innovation actionCoordinateur
1000 Ljubljana
Slovénie