I have used English speech recognition earlier. But it is a completely different experience to be able to control the PC using Norwegian commands!
Story by: Morten Tollefsen - 17.04.2009
Veridict, IBM and MediaLT are collaborating to develop Norwegian speech recognition, and we are now starting to see some very interesting results!
People have different assumptions, some use the mouse a lot, while others prefer the keyboard, and some read with synthetic speech. Yet for many of us, using your voice to control your PC is still science fiction!
Voice recognition is now available for several major world languages. However, the only dictation application currently available for Norwegian is Max Manus’s product for the health sector.
The SMUDI project is researching into the use of different methods for input and output (multimodality). Voice control is a very important focus, and the goal is to develop innovative products and services.
The standard equipment for controlling the PC is the mouse and keyboard. If you are prevented from using these, you need other options. The best option for many is to use their voice. In Norwegian this has so far not been possible. An activity in the SMUDI project is to develop a pilot for Norwegian command control. This will enable you to open programs, make menu selections, give command such as “go to favorites", spell etc., all using your voice. In the first instance text dictation will not be part of the solution, but the goal is PC use with voice only!
The first tests with command control are now being carried out in Stockholm. The results are surprisingly good! Although we have not yet carried out scientific tests with error rates etc., we can say that the system is working beyond our expectations.
Without any training in the system at all, I was able to sit down and control my PC using my voice. Spelling also worked well, and this is a challenge in speech recognition because many sounds are similar.
The Command control system is not yet ready for marketing! Detailed work on the commands and functions still has to be carried out. Moreover, we want to create a very robust produt. This work is time-consuming and costly. Verdict, IBM and MediaLT are now working to finance this part of the development.
We will eventually also test the solution: primarily with users who cannot use a mouse or keyboard. Our ambition is to achieve this in 2009. An invitation will be published on our Web site, and any who are interested in this development can contact us now.
Tips noen om siden