Speech Recognition: There Actually Is An Open Source Solution

by Ostatic Staff - Jul. 21, 2010

If you happen to have followed the path of speech recognition software over the years, you know that it's been a rocky road. Pundits ranging from Ray Kurzweil to Bill Gates, have, at various times, proclaimed speech recognition to be the wave of the computing future, but you probably rely on mice, keyboards and pointing devices to interact with computers more than microphones. Today, there is an interesting query posted on Slashdot, asking whether there is any free, open source software in this area. Well is there?

Speech recognition software has gotten better over the years, but the problem with it remains that it's in the neighborhood of 90 percent accurate--if you train the software to get used to your inflections--and that's not good enough for some people. It is good enough for some, though. I know some journalists who do all their transcriptions with Dragon's proprietary software, or with IBM's ViaVoice. But what are the open source alternatives?

As the commenters on Slashdot note, one of the most robust open source speech recognition solutions comes from Carnegie Mellon University. It's called Sphinx, and we covered it here. You can use Sphinx for straight speech recognition, or integrate it with applications. To find out more about Sphinx, check out this post from Artificial Intelligence and Robotics.

My sense is that Dragon remains the most common speech recognition choice for most people. Sphinx continues to mature, though, and I won't be surprised to see it become adopted by many.