Lewies Blog

I’ve been starting to look into some tools today primarily with speech recognition because now that I’m going to be making some audio files I am going to also post the text along with that audio of what I am actually saying. The reason being is that I would like to be able to search for what I am saying and I would like other people to find what I am saying just by searching through search engines or any kind of search engine that I have on my website.

I started looking up some keywords like "Voice Analyzer/Recognition", "Speech Recognition", "WAV to text", "Wave MP3 Voice Recognition", "Lernout and Hauspie", "Speech Recognition Engine", "Microsoft’s SAPI", "Microsoft Speech SDK", and "Dictation". There are so many keywords that are surrounding this kind of technology.

I went out on the Internet and found a lot of different pieces of software that say that they’ll do the job for you. Some software says that it’ll do it but the interface is really horrible like this “Wave to Text” program. This is the most horrible interface that I’ve ever seen. Then there is another one that is some kind of dictator and I tried it out and it seems like what you are actually supposed to do is look at a written text and then just say what you are reading. That is the opposite of what I want. I want to be able to have the program read what I am saying - or at least write what I am saying.

When I started searching a little bit I learned about Lernout and Hauspie wich is a company that was purchased by someone else (Simtel).

“It understood you just fine provided you had a Belgian accent,” Lernout said. “We were working out a fix for that when our creditors went after that. Until then, we were suggesting that everyone talk like eurotrash to make it work.”

Microsoft was partnering with them or something and they had speech recognition software built into there operating system that listens to what you are saying and carries out commands. Now on top of this, there Microsoft Office has a dictation tool built into it and I didn’t realize this until later tonight. What it does is as I say something, it writes it out.

Problem - It’s not perfect. Now they say as you train the program you’ll get between 85 to 90% and the better it is it can get up to 95% of being correct. I don’t think that I’m up to that 85% just yet. It might be that my voice is not clear, or it might be that I just have not trained this program enough.

I saw that I was training this about two years ago. I still had a profile. I’ve been training it a lot tonight because I figure - two years, my voice has changed. It’s funny to see what it thinks I am saying. In fact it’s actually recording me as I talk into my phone rite now. I have the dictation on at the same time so I’ll just post what it thinks that I’m saying and see what it says up there. Actually it’s some bad words up here that I got to censor or something.

In a nutshell it seems that Microsoft speech dictation might be the way to go. The problem is that it works better if I am talking into the microphone. The thing that I want to try is to talk into a digital voice recorder when I’m not at home talking into this microphone so I got to figure out how toMicrosoft audio software thinks that I said the following:

Up and sexual abuse and schools today, early with speech recognition in his know I’m going to be making some audio files and I want to also pose the text along with an audience of Wenatchee saying the reason being is that I would like to be able to search for what I’m saying. Not like other people to be able to find what I’m saying just by searching for search engines were on the tennis search engine I have a website

well-known was start looking up some key words like was analyzed her voice recognition speech recognition waived text waited three boys recognition off large out-speed spawn speech recognition engine Microsoft’s as a P. I myself speech SDK and a case in their semi differing key words that are surrounding this kind of technology

from highway dory the internet era from a lot different pieces of self a pseudo do the job for you some saw were says I’m still doing with the interfaces new car will review so wage tax program uses the most horrible interface of their perceived as a man was another one man was sold some kind 88 a and I tried out his seems like with your actions supposed to do is locate the written text images say when are we in that seat opposite of what I want I would to be allowed to allow have a program read what I’m saying released write what I’m saying

U.S. are searchable that I learned about one hour and half the wishes of company that was purchased by someone else But Microsoft was partnering with summer something and they have speech recognition software builds into their operating system data and she listens to what you say a curious out commands new on top of this their Microsoft office his he did teaching tool belt into a new realize this until later tonight in windows is as I say something it right now

problem it’s not perfect values saying as you train the program you’ll get between five to 90 percent gain better is seeking give up of many five percent the correct a time I don’t think ULTIMA 85 percent just yet time in might be that might force is not clear what might be that I just have not trained this program enough

guys sell then I was trying this out two years ago I still have profile and the free training in a lot tonight because I figure in two years by voices change from now . Sorry to see where they some say is that the section recording made as I’ve talked to my father now I have a deep a small at the same time soldiers Hulsman thinks that and say is a waste us out of their action way it’s got some bad words to that idea sensor Harsono

(but they’re not sure what seems like from Microsoft’s feature a steep Haitian might be the way deal the problem is a work better if I’m talking into the microphone clothing their want to try is to talk into the digital voice recorder when I arrived home twenty into this with microphone so I cannot figure out how to The

This entry was posted on Friday, April 15th, 2005 at 12:34 am and is filed under Miscellaneous. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.