Transcribo Vox

Identified challenge

Currently, press conferences are transcribed manually by the Information and Press Service of the Government (SIP). Press conferences are transcribed in the spoken language, which can be Luxembourgish, French, German or English.

It should be noted that in the vast majority of cases, press conferences are held in Luxembourgish. In some cases, press conferences are accompanied by an interpreter who provides a French translation of the speeches in Luxembourgish. In these cases, the SIP transcribes and makes available the French version of the interpreter in addition to the original audio/video.

Depending on the constellation of presenters at a press conference, several languages may be used at the same press conference. Each intervention is then transcribed into the spoken language.

The video stream of the press conferences is streamed live (Youtube, Facebook, and possibly on other media: RTL, ...). The recording file is then sent to the SIP in MP4 format, which then publishes it on the platform Youtube. The SIP adds manual subtitling of the spoken language to the final video.

The SIP also takes care of publishing the links to the video(s) as well as the transcription(s) in "DOCX" file via the gouvernement.lu website.

A solution for real-time transcription of audio and/or video files in Luxembourgish, French, English and German is being sought. This solution can then be applied to many other use cases within the public sector. For example, another possible use case is the transcription of certain radio news from Radio RTL Lëtzebuerg and 100.7, which is currently done manually by the SIP on the basis of an audio file, twice a day. The solution will make government communications available in video and audio format to a wider audience.