Tuesday, March 07, 2006
02:59 by FoxTwo This is a special edition of this blog. This is written fully with voice recognition software. After watching the Windows Vista demonstration, I thought I would try my hand at running a voice recognition software to see how it performs.
So far it has been working quite beautifully. The accuracy rate is somewhere around 95%. The best thing about voice recognition software is that nowadays, they do not just work on WordPad or Microsoft Word. As you can see, now, they also work within a Web browser. I personally am using Firefox browser, and as you can see from this paragraph, it works just fine.
At this very moment, I am actually conversing on MSN with a friend using the speech recognition software. It is almost like talking to him naturally. In fact, he doesn't believe that I'm actually talking to my computer.
Does the software actually recognise everything you say? The short answer is, no. There are many instances where I had to go back and correct spelling, because what is heard me say and what I actually wanted to say was different. For example, the word heard and hurt sounds similar to the computer. So, for words like these, I have to go back and correct the spelling myself. Granted, it is nowhere near as impressive as the Windows demonstration, it works well enough for the task at hand.
I don't know whether some of you may remember, but Creative used to bundle text-to-speech and voice recognition software with the Sound Blaster cards. I remember time somewhere back in 1993 (or was it 1994?) where an office colleague of mine brought the bundle to the office. He actually loaded the software onto the office PC and started to play around with the voice recognition software. In fact he was a maniac who retrained the computer to forget English completely and understand only Hokkien. It was damn funny and watching him talk to the computer purely in Hokkien.
Fast forward to 2006. Now we have voice recognition software that actually recognises full and complete sentences and we do not have to talk like robots. In fact this software I'm using recognises the Singlish accent. I do not even have to fake any American or British English accent. But, we are nowhere near Star Trek quality when it comes to voice recognition. You cannot ask the computer to search for something without first starting the Web browser for example. However, it is pretty good at understanding which software you want to launch when you pronounce the software name correctly. You can say things like "Start Microsoft Word" and the computer will automatically launch Microsoft Word. You can even say things like "start Firefox" and it will launch Firefox Web browser. Even obscure software like Gizmo Project (VOIP software) is recognised by the speech recognition software and it will launch accordingly, provided you pronounced the name correctly. you do not have to go through the whole process of saying "start, program files, Firefox".
So the final verdict is, voice recognition software is fun to play around with but not entirely useful for business applications. You will probably pull your hair out in frustration in trying to get it to understand your particular style of speaking. You can overcome this by training the software's accuracy using the accuracy trainer. The more you train it the better it is able to understand how you speak.
Maybe that's what I should be doing: keep training and retraining the software to recognise my style of speaking so that in the future, I can truly have a hands-off experience when I start to write my e-mails or my documents. In fact I am hoping that I can soon write my blog totally hands-free.
02:59 by FoxTwo This is a special edition of this blog. This is written fully with voice recognition software. After watching the Windows Vista demonstration, I thought I would try my hand at running a voice recognition software to see how it performs.
So far it has been working quite beautifully. The accuracy rate is somewhere around 95%. The best thing about voice recognition software is that nowadays, they do not just work on WordPad or Microsoft Word. As you can see, now, they also work within a Web browser. I personally am using Firefox browser, and as you can see from this paragraph, it works just fine.
At this very moment, I am actually conversing on MSN with a friend using the speech recognition software. It is almost like talking to him naturally. In fact, he doesn't believe that I'm actually talking to my computer.
Does the software actually recognise everything you say? The short answer is, no. There are many instances where I had to go back and correct spelling, because what is heard me say and what I actually wanted to say was different. For example, the word heard and hurt sounds similar to the computer. So, for words like these, I have to go back and correct the spelling myself. Granted, it is nowhere near as impressive as the Windows demonstration, it works well enough for the task at hand.
I don't know whether some of you may remember, but Creative used to bundle text-to-speech and voice recognition software with the Sound Blaster cards. I remember time somewhere back in 1993 (or was it 1994?) where an office colleague of mine brought the bundle to the office. He actually loaded the software onto the office PC and started to play around with the voice recognition software. In fact he was a maniac who retrained the computer to forget English completely and understand only Hokkien. It was damn funny and watching him talk to the computer purely in Hokkien.
Fast forward to 2006. Now we have voice recognition software that actually recognises full and complete sentences and we do not have to talk like robots. In fact this software I'm using recognises the Singlish accent. I do not even have to fake any American or British English accent. But, we are nowhere near Star Trek quality when it comes to voice recognition. You cannot ask the computer to search for something without first starting the Web browser for example. However, it is pretty good at understanding which software you want to launch when you pronounce the software name correctly. You can say things like "Start Microsoft Word" and the computer will automatically launch Microsoft Word. You can even say things like "start Firefox" and it will launch Firefox Web browser. Even obscure software like Gizmo Project (VOIP software) is recognised by the speech recognition software and it will launch accordingly, provided you pronounced the name correctly. you do not have to go through the whole process of saying "start, program files, Firefox".
So the final verdict is, voice recognition software is fun to play around with but not entirely useful for business applications. You will probably pull your hair out in frustration in trying to get it to understand your particular style of speaking. You can overcome this by training the software's accuracy using the accuracy trainer. The more you train it the better it is able to understand how you speak.
Maybe that's what I should be doing: keep training and retraining the software to recognise my style of speaking so that in the future, I can truly have a hands-off experience when I start to write my e-mails or my documents. In fact I am hoping that I can soon write my blog totally hands-free.
Tags:
