Speech recognition, circa 2005

I have always been interested in speech recognition technology this is the first time in more than twenty years that I have tried using speech recognition and the first time I have never tried using it for dictation in 1984 and three on science center I use the system that recognizes about ten different words come down (a fast slow and selling to control of the cool jury plane of the game now than 2500 items covered the Windows XP speech recognition system enables it to them as they speak my speeches captured into Microsoft word and I am not looking at the screen as I speak so that I can be surprised at the end with the accuracy or inaccuracy with which the system to raise my voice publicity place recognition is very tricky because each individual speaks with a different accents were using different languages they are words that sound similar to each other better still very differently and have very different meeting as Sylvia will be quite fascinating to me to see how accurate the result of this speech to text conversion is in general I have little faith in the systems because for all of their computing power computers is still lack the ability to place the context into the various pattern matching algorithm is that enabled speech recognition video interpretation sophisticated motion detection and other actions that might seem quite trivial to human being or brains were very differently and confident that in the future these issues with the interest rate for the time being placed back issues are not high now I’ll stop speaking and read that would have spoken so far and post this to my wallet so that you can see the level of technology for yourself

Comments

neal said…
seems mostly coherent...

Although is it safe to say that it recognized most of your pluralizations as "is" instead of a pluralization?

I find that quite interesting...
Sassan Sanei said…
I have to say that it did better than I expected, but obviously there are still a lot of faults.

For example, "In 1984 at the Ontario Science Centre I used a sytem..." turned out as "in 1984 and three on science center I use the system..." The rest of that sentence does not make a whole lot of sense either - I don't know how it got "cool jury plane" and I never said 2500. It did recognize "Windows XP" and "Microsoft" easily enough, though. :)

And it has no concept of the start and end of a sentence.

I think we are at the point where speech recognition can be used reasonably reliably for commands and navigation, i.e. choosing from a fixed number of options, but for dictation such as I tried today we are a long way away from reliable.

Still it is fascinating to me how much the technology has advanced.

I am going to read more of the "training" passages and see how the quality improves.

Thanks for your comments.

Popular Posts