Speech recognition software allows you to operate a computer by voice. Speech- operated computing has developed significantly since it first emerged on the market – when it was necessary to pause between each word, and made a lot more mistakes than it does today. These days, speech recognition actually works best when longer phrases are spoken, and accuracy can be as high as 98 to 99%. A range of computer tasks can be done by voice, including dictation, punctuation, formatting, navigation, correction of recognition errors, commands such as “print document”, “save document” etc (using either built-in commands or custom commands/macros), editing text, opening applications and documents, writing and managing emails, browsing the web and editing spreadsheets.
Dragon NaturallySpeaking is the market leader in speech-recognition computing for PCs, and includes an Australian accent model. Other speech recognition products available include ViaVoice and Apple Macintosh’s iListen. Vista, Microsoft’s new operating system, will include a speech recognition system. It remains to be seen whether Vista’s system will work well with Australian accents.
Dragon NaturallySpeaking is the market leader in speech-recognition computing for PCs, and includes an Australian accent model. Other speech recognition products available include ViaVoice and Apple Macintosh’s iListen. Vista, Microsoft’s new operating system, will include a speech recognition system. It remains to be seen whether Vista’s system will work well with Australian accents.
Speech recognition software works by picking up sound waves via the microphone in analogue form, which are then converted to digital information by the soundcard. Speech recognition software matches sounds to words using a vocabulary, and a language model that assigns a statistical probability to a particular word occurring in speech, as well as the probability of words occurring in a particular context, that is groups of two and three words occurring together. For example, while the phrases “whirled tour” and “world tour” sound the same, the software will choose the latter because this combination is more common in speech.
Recent versions of Dragon NaturallySpeaking also assess the probability of groups of four words occurring together. This is why you will get better recognition if you speak in longer phrases or sentences. Through correct use of the software, the vocabulary and language model will adapt to the user’s speech. Correction of recognition errors is one way of improving recognition over time.
These pages provide advice on how you can get the most out of speech recognition software, particularly for those ‘high-end’ users who undertake complex computing tasks, and/or have a medical condition or disability that makes it difficult to operate a computer by hand.
While you’re here…
While you’re here, how about helping us out with a donation?
We’re a really small organisation doing a really big job. We give people with RSI across Australia the info they need to get the right treatments, navigate the worker’s compensation system and better manage daily life.
Every little bit helps – so make a donation here to help us as Australia’s only RSI support organisation.