Machine Interpreting - Demo (v 1.1)

Select the right audio input channel (Stereo Mix on Windows). Click on MIC and on video PLAY

On air.

No speech was detected. You may need to adjust your audio/microphone settings.

Click the "Allow" button above to enable your microphone.

Permission to use microphone was denied.

Permission to use microphone is blocked. To change, go to chrome://settings/contentExceptions#media-stream

Machine Interpreting is not supported by this browser. Upgrade to a recent version of Chrome.

                

Welcome! I am a digital interpreter. While I am waiting for the speech to begin, I'd like to tell you something about me...

I have been developed in the context of academic research aiming at studying machine interpreting (also called speech-to-speech or speech-to-text translation) by Claudio Fantinuoli at the University of Mainz.

The project targets conference interpreting, i.e. the simultaneous and consecutive modalities. For now I target only formal speeches.

At the moment I work better in short consecutive mode, but I can also work on continuous speeches (like the ones on the right) approaching them in a slightly different way as human conference interpreters do. Well, I am not human...

Do not expect from me a human-like performance. I am definitely not there. However, I will improve in the future. For example, I will soon be able to reduce my decalage (latency), since my linguistic knowledge will increase.

Claudio and others will use me, modify me, upgrade me, to understand - from a communicative perspective - how good I am, what are my limits, which speeches can I interpret best, and foremost to shed light on how special you, your brain, your language capabilities, are.


Roadmap

  • Version 1: baseline approach
  • Version 2: linguistic-aware sequencing of continuous speech
  • Version 3: real-time sentiment analysis and framing

Read here to know how to use the demo.



You may use other videos with the supported speech typology: formal speech, slow delivery speed with pauses (this will change in version 2.0).