Experimenting with new tools for real-time subtitling

Remember that experiment we did on real-time subtitling? In this post we briefly reported how a student in her final year of interpreting studies, together with a recent graduate, tested the full potential of Dragon Naturally Speaking to produce speech-based real-time subtitling.

Given the very positive results we obtained from that first try, we decided to push it a little further and, in the past months, we experimented with new ways to improve that system.

In collaboration with our departmental computer expert Gabriele Carioli, we finally end up with a new tool, which substantially enhances subtitling performance. Besides interacting with the speech recognition software, this new tool enables users to review subtitles in real-time mode, before they are screened.

The process is in itself straightforward: subtitles are automatically produced by associating interpreting with a speech recognition software, while a reviewer edits any mistakes found and sends the final output to a projector screen immediately afterwards.

On Sunday 9th October, our re-speakers Giulia Salani and Alessia Pollice, supervised by Claudia Lecci, tested this new subtitling system at Diego Fabbri Theatre in Forlì during a speech by Joseph Stiglitz (recipient of the Nobel Prize in Economics in 2001). Results were quite promising, so watch this space for updates since we will keep improving!

Want to know how it works? Watch the video below!