Imagine a software using the newest technologies for translation, voiceover and other translation tasks for almost all languages spoken around the world. If a well-designed platform would be available, this could save editors thousands of hours each year, freeing resources for deeper reporting. Imagine further that these technologies are moving from research to the real world, as they are getting better quickly. This includes cross-translation from a large number of languages, including those that are “far away” from each other, such as English and Chinese, or Farsi and German.
New Funding from Google Digital News Initiative
Recently Deutsche Welle Research and Cooperation Projects has successfully obtained funding from the Google “Digital News Initiative” (DNI) Program for a new effort to make translation considerably easier for newsrooms. The name of the project: news.bridge. This extended project will build on the outcomes of yet another, smaller, 6-month prototype DNI project called speech.media, which ends in July 2017.
Extend Usage of Innovative Transcription and Translation Tools
news.bridge will develop a platform for a seamless workflow of automated transcription, translation, voiceover and summarization for currently over 8o different languages, made possible through the use of already existing tools such as Google Translate and IBM Watson. The combination of these two platforms (and other similar tools) makes the project particularly powerful.
Translation and Automated Speech Recognition Getting Better
“It’s time to re-calibrate your earlier impressions from such tools, like Google Translate from just two years ago. At the time these tools produced readable output, but had a fair amount of mistakes, which made them difficult to use in newsroom environments. In the past 18 months language and translation software made incredible progress in terms of quality”, says Peggy van der Kreeft, who is the expert for language technologies in the Deutsche Welle research team.
news.bridge will provide specific workflows combined with editorial control to ensure quality of automated output. “This will allow us to render almost all of the Deutsche Welle broadcast content in virtually any language in the form of subtitles, transcribed text or voiceover”, adds van der Kreeft, who will manage the project on behalf of DW.
Partners in the Project
The project will be launched later in 2017 and is coordinated by Deutsche Welle, who is also the primary user partner and beta tester. LETA, the Latvian News Agency, is platform developer. Le Mans University provides transcription and translation enhancement modules. Last but not least, Priberam, a Lisbon-based natural language processing SME, develops a summarization tool. All this combined is expected to result in a powerful, robust , digital media translation tool.
We are moving towards faster workflows, higher quality and more options for newsrooms. Stay tuned.