The Project

What is ELITR

The project “European Live Translator” is supported by the EU under the Horizon 2020 programme, and coordinated by Ondřej Bojar, Charles University, Prague.

It runs from January 1st, 2019 to December 31st, 2021.

The partners:

The goal of the ELITR project is to remove the language barrier in communication within and among European citizens, companies, institutes and organizations at large assemblies like conferences, in smaller live discussions like workshops and in discussions held over long distance like formal or informal online meetings. Within the project, we will create automatic subtitling system of live meetings and conference presentations and provide the system of spoken language translation (interpreting). Furthermore, the project will design and implement automatic minuting, i.e. create structured summaries from automatic transcript of a discussion. We also aim to advance the state of the art to deliver machine translation of high-quality at the level of whole documents and to scale up to multilingual settings with dozens of source and/or target languages. The technologies will be tested by the Supreme Audit Office of the Czech Republic and alfaview®, a German online conferencing system.

Keywords:

  • Machine translation
  • Multilingualism, language diversity
  • Automated translation
  • Natural language processing
  • Service oriented architectures

Objectives

Specifically, the ELITR project targets the following five objectives:

  • Advancement and application of ASR and SLT in two challenging settings:
  1. Face-to-face conferences (interpreting official speeches and workshop-style discussions), and
  2. Remote conferences (interpreting discussions held over a on-line platform).
  • Advancement of machine translation to achieve:
  1. Robust and efficient systems for machine translation with many source and/or target languages, and
  2. Coherent translation of larger documents and dialogues.
  • Advancement of speech understanding in a highly practical application:
  1. Automatic minuting, i.e. summarizing meetings in the form of structured minutes from speech.

Challenges

The practical setting of our use cases however remains very challenging with the current technology readiness level for a number of reasons:

  • Non-native speech has to be handled gracefully in EUROSAI workshops and on-line meetings.
  • Very high SLT quality is needed, underpinning the importance of the EUROSAI Congress.
  • The challenging domain of auditing demands trustworthy translation outputs and meaning preservation of the precise formulations, at the level of whole documents.
  • The immense multilinguality (up to 1806 language pairs of EUROSAI languages) lies beyond the technical limits of current approaches to MT.
  • Success in the proposed automatic minuting could cause a technological breakthrough in live and remote meetings.