Version 0

Lecture: FLOSS speech recognition: Where are we?

And why aren't we there yet?


In this talk I will try to answer why we don't yet have a free version of Apples Siri or Nuance's Dragon Naturally Speaking and what we can do about it.

"Speech recognition" is a very wide field. On the one hand, there are the command and control applications which recognize stuff like "Open browser!" "Next window!". On the other hand, there are large vocabulary speech recognition setups that don't restrict the user to such a narrow band of sentences, affording completely different use cases: dictating texts, for example. More recently, electronic personal assistants like Siri or Google Now have established themselves as helpful features instead of the convoluted tech-demos their predecessors always were.

While command and control applications are comparatively easy to build using free open source technology, large vocabulary systems have remained firmly in the realm of commercial applications. As speech recognition is becoming more and more important there is really no reason why we can not or should not compete with the best of them.

In this talk I will lay out the difficulties in building advanced speech recognition systems and why these have historically hit open source initiatives the hardest. I will present the current state of open source speech recognition and outline concrete steps on how we can navigate apparent roadblocks to build truly open, competitive speech recognition systems.


Day: 2013-07-14
Start time: 15:30
Duration: 00:30
Track: New Ideas




Click here to let us know how you liked this event.