Speech recognition software can be a godsend for busy executives whose job demands lots of writing, the disabled as well as for people who just aren't very good at typing. Nuance’s Dragon Naturally Speaking is the most well-known and most popular speech-to-text program and it has a long history.
Dragon Naturally Speaking aims to reduce your keyboard usage by letting you issue voice commands to control Windows as well as dictate documents; which automatically converts your words into text as you speak.
Installation is simple, although you’ll need about 2GB of hard disk space on your main Windows partition regardless of what operating system you use as from window xp to most recent OS. Having activated the software online, the next step is to spend a few minutes training Dragon to recognise your voice (multiple profiles are allowed for different users). Heavy accents can be an issue, but there are presets to help you out of that.
You can also transcribe from digital audio files or by plugging in a voice recorder via USB pot.
Besides dictating text and formatting instructions (such as “comma” or “delete back two words”), you can also use NaturallySpeaking to control your PC. A similar feature is built into Windows 7, but having both together is useful. For example, while working in Word you can save or open a file without reaching for the mouse. There’s a wide range of commands for Windows and Office apps, and handy new smart shortcuts allow complex operation such as “Search Wikipedia for autodidact.” New commands can be set up to enter preset text and/or graphics, but not to control software or system functions.
Software that actually understands what you say is still science fiction, but DNS uses a whole range of tricks to detect words correctly. While using dragon, you may have one mistake every few sentences, which, considering the complexity of speech and how many words sound alike, is impressive. Your results may differ, but should continue to improve over time. NaturallySpeaking is three times faster than anybody can attain while typing, and it does it with optimum accuracy.
The software needs some basic information before you start working with it. It needs to know what region of the world you live in and any accent you might speak with, as well as your age bracket. This basic data gives it a few pointers to help it understand your voice. As part of the short induction process, you have to read out a few sample sentences to further train the software to your voice.
The setup process takes under a quarter of an hour. With that completed, you can let the software scan your "My Documents folder" so it can analyze your writing style and also pick up commonly typed bits of information such as the email addresses you use, to help it improve the accuracy of its speech recognition. How long that takes will vary depending on in the size of the folder. If, like us, you have a lot of content in that folder produced by other people, it might be worth skipping that part of the process as I did when testing out the app.
Accuracy was surprisingly good, and I really like the way you can make corrections. Simply saying ‘correct’ and what has been incorrectly transcribed, or saying ‘correct that’ calls up a menu from which you can choose the words to be corrected, spelling out the correct words, convert lower case text into capital letters, or simply type over the incorrectly transcribed text to teach the program what it should have done.
Possibly the most imperative step in using voice recognition software is creating a voice profile. This allows the program to adapt to your voice, your accent and your verbal idiosyncrasies so it can better understand what you say. So it should come as no surprise that DNS does an excellent job of creating a profile for each user and providing the tools to make the program more accurate.
If you say a command and Dragon NaturallySpeaking thinks there is more than one option on the screen that applies to what you said, it numbers the possible responses. All you have to do is say the number corresponding to the command you meant. This is an exceptionally good design feature because it allows you to navigate complex sites and menus without the program bogging down.
Dragon Naturally Speaking was accurate enough and fast enough to keep up with use speaking at regular conversational speeds; meaning you don’t have to talk slower or enunciate more than you normally do when using the product (unless, of course, you are a chronic mumbler). In fact, it actually seemed to perform better at normal speaking pace than when we slowed down and enunciated. I noticed that at standard conversational speeds, the program occasionally missed some short words altogether, like “the” and “to,” so you may still need to emphasize those shorter connecting words a little more than normal.
Dragon Naturally Speaking also handles punctuation commands nicely, so that there was very little editing needing to be done before the document was grammatically correct. In fact, if you seldom use more than the standard period, comma, question mark and quotation mark, you’ll almost never need to edit your punctuation commands.
Dragon Naturally Speaking aims to reduce your keyboard usage by letting you issue voice commands to control Windows as well as dictate documents; which automatically converts your words into text as you speak.
Installation is simple, although you’ll need about 2GB of hard disk space on your main Windows partition regardless of what operating system you use as from window xp to most recent OS. Having activated the software online, the next step is to spend a few minutes training Dragon to recognise your voice (multiple profiles are allowed for different users). Heavy accents can be an issue, but there are presets to help you out of that.
You can also transcribe from digital audio files or by plugging in a voice recorder via USB pot.
Besides dictating text and formatting instructions (such as “comma” or “delete back two words”), you can also use NaturallySpeaking to control your PC. A similar feature is built into Windows 7, but having both together is useful. For example, while working in Word you can save or open a file without reaching for the mouse. There’s a wide range of commands for Windows and Office apps, and handy new smart shortcuts allow complex operation such as “Search Wikipedia for autodidact.” New commands can be set up to enter preset text and/or graphics, but not to control software or system functions.
Software that actually understands what you say is still science fiction, but DNS uses a whole range of tricks to detect words correctly. While using dragon, you may have one mistake every few sentences, which, considering the complexity of speech and how many words sound alike, is impressive. Your results may differ, but should continue to improve over time. NaturallySpeaking is three times faster than anybody can attain while typing, and it does it with optimum accuracy.
The software needs some basic information before you start working with it. It needs to know what region of the world you live in and any accent you might speak with, as well as your age bracket. This basic data gives it a few pointers to help it understand your voice. As part of the short induction process, you have to read out a few sample sentences to further train the software to your voice.
The setup process takes under a quarter of an hour. With that completed, you can let the software scan your "My Documents folder" so it can analyze your writing style and also pick up commonly typed bits of information such as the email addresses you use, to help it improve the accuracy of its speech recognition. How long that takes will vary depending on in the size of the folder. If, like us, you have a lot of content in that folder produced by other people, it might be worth skipping that part of the process as I did when testing out the app.
Accuracy was surprisingly good, and I really like the way you can make corrections. Simply saying ‘correct’ and what has been incorrectly transcribed, or saying ‘correct that’ calls up a menu from which you can choose the words to be corrected, spelling out the correct words, convert lower case text into capital letters, or simply type over the incorrectly transcribed text to teach the program what it should have done.
Possibly the most imperative step in using voice recognition software is creating a voice profile. This allows the program to adapt to your voice, your accent and your verbal idiosyncrasies so it can better understand what you say. So it should come as no surprise that DNS does an excellent job of creating a profile for each user and providing the tools to make the program more accurate.
If you say a command and Dragon NaturallySpeaking thinks there is more than one option on the screen that applies to what you said, it numbers the possible responses. All you have to do is say the number corresponding to the command you meant. This is an exceptionally good design feature because it allows you to navigate complex sites and menus without the program bogging down.
Dragon Naturally Speaking was accurate enough and fast enough to keep up with use speaking at regular conversational speeds; meaning you don’t have to talk slower or enunciate more than you normally do when using the product (unless, of course, you are a chronic mumbler). In fact, it actually seemed to perform better at normal speaking pace than when we slowed down and enunciated. I noticed that at standard conversational speeds, the program occasionally missed some short words altogether, like “the” and “to,” so you may still need to emphasize those shorter connecting words a little more than normal.
Dragon Naturally Speaking also handles punctuation commands nicely, so that there was very little editing needing to be done before the document was grammatically correct. In fact, if you seldom use more than the standard period, comma, question mark and quotation mark, you’ll almost never need to edit your punctuation commands.
