The SpeDial partners. From left to right: Dominique Boucher (Nu Echo), Fernando Batista (INESC-ID), Katerina Louka (VoiceWeb), Isabel Trancoso (INESC-ID), Joakim Gustafson (KTH), and coordinator Alexandros Potamianos (National Technical University of Athens).
Two weeks ago in Luxembourg, the international team pictured above presented the final results of the SpeDial project to reviewers of the European Commission. The project, which lasted two years, was funded by the EU 7th Framework Programme (See here for a short description of the project).
The presentation went very well and we received very positive feedback from the reviewers: “Great Project”, “Fantastic Team”, “Wonderful Execution”, “Concrete Work based on Deep Studies”, “Fantastic Collaboration”.
This project was the result of a tight collaboration between both SMEs (VoiceWeb and Nu Echo) and the research partners (KTH, INESC-ID, Athena Research and Innovation Center, and TSI). We owe a big thank to Alexandros Potamianos, who did an exceptional job coordinating this project involving both European and North American partners. It was not easy to get a non-European partner in the consortium; but Nu Echo managed to make it happen.
A real opportunity of this project, other than the international exposure and new research, was that it enabled us to enhance some of our proprietary technologies which we use for automated detection of perfformance issues (potentials for improvement). During the final review, I showcased a tuning project Nu Echo worked on a couple years ago that required (at that time) a fair amount of call analysis to pinpoint the cause of the most critical performance problems. Using the new technologies we have developed, detecting those issues took a fraction of the time; we now simply run a report which happens in less than a minute! Of course, some additional time is required to validate hot-spots, but it’s nothing compared to listening to hundreds of calls, analyzing each issue manually, and trying to find the common denominator of what is causing the issue.
In the end, what does all of this mean for our clients? An improved ability to find tuning opportunities more efficiently in speech applications, resulting in less time spent analyzing data and more time doing actual tuning work. That is especially important for projects on tight schedule and budget.