Wow, so many questions! I shall try and answer them as best I can!
“I've tried to use speech recognition software at schools before, with mixed results. This was a few years ago, mind, so maybe you can update me a bit if things have moved on by now.”
Dragon NaturallySpeaking has improved massively over the 11 years or so of it’s life and the latest version of the product, version 10, came out last September, with an update recently released to patch a few minor bugs, and enable 64-bit compatibility with Microsoft Vista. The recognition capabilities of this product are absolutely outstanding. There were other speech recognition products available to consumers over the years such as ViaVoice, but these have now gone out of production.
“We were working with SEN pupils, trying to set up speech recgonition so they could simply talk into a microphone instead of having to type - useful for pupils who couldn't type or maybe with dyslexia or similar. The software needed to be "trained" to recognise a particular voice (don't know if this is still the case).”
All speech recognition software needs some degree of training, and of course so does the user. Although the current version of the product allows you to select a "no training" option, it is better to do some degree of training so that they product learns the subtleties of the voice. Dragon has to cope with the various ways that we say different words, (think of the differences in the way that a Northerner would say "book", or "grass" against a Southerner!). Also the person needs to have a consistent voice, so that the program has a reasonable voice pattern against which to match words. With children their voice seems to become more consistent between the ages of 8 and 10, but this is not an empirical statement as it really depends on the individual.
“We found the voice training process to be rather long for a child and that the child needed a certain level of literacy in the first place to be able to read the example text.”
The training required for the product these days is about 5 to 10 minutes, and of course requires that the child is able to read, unless of course a teacher is able to whisper the training text into their ear during the training process.
“Voice also takes a fair bit of data (the settings file for a particular user was around 100MB, I seem to remember), which is a problem to deal with in a school where pupils move from workstation to workstation.
How well does Dragon Naturally Speaking work over a network now? How does it store its data files, how big are they - does it still need voice training? Is there a proper network installer for the software - an MSI file we can distribute around via GPO? Is there a server based option that we can submit a sound file to and have it return text, so pupils could use any machine that had a microphone?”
I would love to be able to give you a definitive statement on this, but it really depends on the version of the product that you are using. It also depends on the recommendations of the company that you are working with to install into a school setup. I really wouldn't recommend a school goes it alone with a product like Dragon NaturallySpeaking, simply because it is a program that benefits from the users being trained properly. At the end of this note I will add a list of companies that are familiar with working with schools of all types.
“How well does the software cope with laptop microphones these days? I know we had issues getting our software to work with laptops because laptop manufacturers generally skimp a bit on the quality of the microphone input.”
Speech recognition software necessarily requires a good quality microphone, and as you say the microphone supplied internally on a laptop just isn't up to the job. We would always recommend the use of USB microphones. Again, this is where working with a company that is familiar with this kind of environment would reap dividends for all users, because they can make appropriate suggestions.
“Does your software work with sound files recorded from portable recording devices?”
The preferred version of the product and above will work with .WAV,.MP3 and .WMA files of a single user’s voice, as long as a users profile has been trained previously.
“Any experience with your software on netbooks?”
I have it running on my Samsung NC10 netbook, and Dragon seems to run okay with some fairly minor restrictions,(which would possibly be overcome if I upgraded the memory to 2 GB).
“Does your software only run on Windows?”
yes Dragon NaturallySpeaking only works on Windows XP 32-bit, and also Windows Vista 32-bit and 64-bit, and we see no issues currently with Windows 7, (although this is not currently supported whilst the OS is in beta).
If you require software for the Macintosh, then I would recommend MacSpeech. MacSpeech is not one of Nuance's products, but is based on the Dragon NaturallySpeaking speech recognition engine.
I hope that this all helps?!
Gary
Companies that specialise in Dragon in educational environments: --
Pugh Computers
Citnexus
Microlink
Barry Bennett
Iansyst
