Speech recognition with a disability
Speech recognition technology is powerful stuff. I have always thought that substantial benefits could await people with a disability. Those with physical disabilities or vision impairment could particularly benefit.
But what if speech problems are part of your disability? That describes me. I had imagined my speech would reduce the benefits of speech recognition technology. So it was a revelation when I discovered that the software improved my life.
Conversations with my computer
I have a neurological disorder which affects my motor skills. I cannot type an email without spending excessive time, energy and frustration. My condition renders my speech slow and slurred. I sound like I'm on a drunken bender. It's actually a capital reason to get on the sauce.
I could not find any advice about how speech recognition software would work with my swampy speech. I made a conscious decision to try (buy) the software though. I bought Dragon NaturallySpeaking and started having conversations with my computer. I chose it because it was widely acknowledged as the best speech recognition software available. I thought that if it worked it could mean a decline in the function of my less-used wrists and fingers. But the positive would be that it could help maintain the quality of my speech. It was a trade-off I was willing to take.
I've been using NaturallySpeaking for 12 years now. I am confident that the trade-off was worthwhile. I communicate well, whether by printed word or speech. I now can write and work at a speed previously unimaginable. Recognition (wreck-ignition?) errors are regular, but even after making corrections I save time. It's not so much that I'm working quicker, but that I can work at all.
I use the software primarily for work which would otherwise be impossible. This is in addition to email contact and communicating ideas. I also use the software for navigating the web and other applications. Training specific words, dictation shortcuts and computer commands in natural language are handy functions. I'm not a power-user but I have enough knowledge to do what I require.
The software requires more initial training for my voice than the average person. Training involves dictating long passages from set texts. I also need to be vigilant in correcting recognition mistakes.
After using NaturallySpeaking for over a decade I now have plenty of helpful hints for new users. My hints include:
- Keep the microphone or headset in a consistent position. Say "microphone off" when not using it
- Backup your speech files regularly. Losing them can be heartbreaking
- Proofread carefully. It is a good general rule for all writing. But it is especially important because the software often picks up on breathing and assorted grunts
- Teach the software to recognise swear words at your own peril. It may seem cool in emails to friends, but an unexpected four letter word does not curry the favour of a lecturer marking an assignment
- Do not use speech recognition technology while drunk. This rule particularly applies when it seems imperative to email or post your opinions
- User files are specific to your voice. Do not let other people use your files, Guard them like your toothbrush.
Control impulse to update
Another important tip is to control your impulse to regularly update your software. I am sure the software makers do not want me to say this. But there are only small improvements with each new version.
More importantly, you can only migrate your vocabulary, not your speech files to a new version. It means you have to complete the training sessions all over again. I have found that newer versions have only marginally better speech recognition. I update only when compelled to do so. Improvements in the software must justify the hard work and expense.
Speech recognition software has been invaluable for me. It gives me an opportunity to write and communicate. It is not cheap at around $300 for a good-quality current version. I also cannot guarantee it will work well for everyone. But it is well worth trying.