Obliterating Language Boundaries

The Heidelberg speech technology company EML European Media Laboratory GmbH participates in the new EU funded research project “TransLectures”. The aim of the project is to provide user-friendly and cost-effective solutions for transcription and translation of educational video lectures provided in online repositories via the Internet.

Alongside EML, the partners of the project are universities, research institutions and companies from Spain, France, Slovenia, the UK, and Germany. The project is co-ordinated by the Technical University of Valencia. “TransLectures” is being funded by the EU until October 2014 with 3,1 million Euro.

Online multimedia repositories have been rapidly growing in recent years. A well-known example of this is the VideoLectures.NET web portal, a free and open access educational video lectures repository that provides with over 14,000 educational videos a huge knowledge source. However, the lectures are hardly searchable and available only in the original language. “TransLectures” is working on technical solutions for the automatic transcription of video content and translation into various languages. Content thus becomes easier searchable, classifiable and analyzable. Furthermore, it’s accessible to an international audience as well as to people with disabilities.

The project aims at exploiting the full potential of automatic speech recognition and machine translation technology to provide accurate enough results for video lectures. The focus of transLectures is on three scientific and technological objectives: Firstly general-purpose models will be adapted with lecture specific models e.g. from accompanying foils. Secondly, the system will allow intelligent interactions with the users. Finally, the system will be integrated into the open-source “Matterhorn” platform to enable real-life evaluation.

EML’s contribution to “TransLectures” is to provide the infrastructure for automatic transcription of video lectures and talks from the VideoLectures.NET and poliMedia web portals. To improve speech recognition EMLs focus is on fast methods to adapt language models on video content automatically. Besides that, EML speech technology engineers develop language models for German, English and Spanish.

The transcription system is to be tested in English and Slovenian within VideoLectures.NET and in Spanish within the poliMedia web portal. Translations shall proceed from Spanish and Slovenian into English and from English into French, German, Slovenian and Spanish.

Project Website: http://translectures.eu/

EML European Media Laboratory GmbH
The EML European Media Laboratory GmbH is a private IT enterprise established by Klaus Tschira, one of the founders of the SAP AG software company. In accordance with its motto “Experience IT – Intuitive Technology” EML is successfully pursuing research and development in the fields of human-computer interaction and automatic speech processing.
Press Contact
Dr. Peter Saueressig
Public Relations
EML European Media Laboratory GmbH
Schloss-Wolfsbrunnenweg 35
D-69118 Heidelberg
Tel.: +49-(0)6221-533245
Fax: +49-(0)6221-533298
Email: saueressig@eml.org