Lately I've been curious about using speech recognition to dictate content. I've heard people talking about using speech recognition to dictate in generally their experiences some pretty positive. I certainly need some way of speeding up my creation of text. When I'm trying to type text I often seem to get stuck ones and a lot of time dawdling. My hope is that by using speech recognition I can speed up the process of content creation because my thoughts flow better when I'm speaking instead of talking. I tend to be very garrulous, even loquacious, so when I'm talking, the words simply flow. When I'm trying to type, somehow I spend a lot of time staring at the screen.
I started out by trying to use the speech recognition facility built into Windows 7 and Vista. I figured that since it is free, it was at least worth a try. Unfortunately though, as far as I can tell from my experiments, it really isn't ready for prime time. Even after extensive training with my voice and after trying several different handsets, it still made many mistakes, usually several per sentence. Moreover, correcting mistakes was a real trial. Even after repeated attempts, it typically failed to get the correct word or phrase. When I gave up on trying to speak the word and tried to spell it instead, it had terrible problems recognizing the letters that I was dictating, failing to differentiate between letters with similar sounds, or between letters and numbers that sounded similar. Another problem I had was that the built-in speech recognition seemed to be incompatible with anything but the Microsoft Office suite. I had really looked forward to using speech recognition to dictate e-mail and blog entries and documents in Google Docs, but basically it was almost impossible to dictate text in any of the browsers I tried, including Microsoft's own Internet Explorer. Sometimes I was able to enter a few words, but then we service simply stop working until I closed the browser and reopened it. The built-in speech recognition might be fine if all you want to do is work with Microsoft office and perhaps spend more time training it were your voice than I did.
One of my friends suggested Dragon NaturallySpeaking. The friend who recommended it was very
enthusiastic, and based on his recommendation I went ahead and bought.I just installed it about half an hour ago on this machine, and so far I am really impressed. I only had to spend a few minutes training it and now I am dictating this blog post. Overall it seems to make fewer mistakes than the Windows built-in speech recognition even with much less training, and the mistakes it makes tend to be a little less incomprehensible and easier to correct. It sometimes seems to drop words toward the end of a long passage and sometimes misinterprets my navigation instructions as text to be entered. Overall though, my initial impressions are pretty positive. I'm curious to see what will happen as I continue using it and it gets better at recognizing my voice. Entering this paragraph already seems to be lot smoother then entering the first paragraph up above. I certainly hope that this will let me post more often to my blog. We'll see how it goes.
UPDATE: I have been using Dragon NaturallySpeaking intensively now for several days. Clearly, it seems to work best with the Microsoft Office suite. I find that it does work with Google Docs, Gmail, and Blogger, though with a few minor glitches. At least when I am using Chrome, when I want to correct something in Gmail or blogger, the correction menu often doesn't appear, or it does appear, but then disappears again almost instantaneously. For some reason, I don't have that problem in Google Docs. However, in Google Docs, the replacement of words following a correction seems a little glitchy. If I correct a word, sometimes it fails to delete the first or last letter of the word being replaced and overwrites a letter or space adjoining the correction. That is only an occasional problem. I have just dictated this entire paragraph was almost no problem at all and only a few corrections to the transcription.
One other issue that I've noticed with Dragon NaturallySpeaking is that it sometimes seems to have a problem recognizing commands as a command as opposed to text to be entered. Maybe it's just me and seems to have a real problem recognizing me saying "go to end of line" and often mistakes that as text to be entered. Just now, it entered that text as " boat go off-line". I have similar problems with certain other commands but by no means all commands. Conversely, it sometimes misinterprets text to be dictated as a command. Again, it may be me, but somehow something that I am saying often triggers it to begin reading text back to me. I haven't figured out quite what the magic word is that this triggering this. Overall though, these are very minor glitches and not too much of a distraction. Moreover, they seem to be becoming less common as Dragon NaturallySpeaking becomes more and more used to my voice. Probably I just need to do some additional training focused on the specific commands that I am having trouble with.
I do like the experience of dictating text. For some reason, I find it easier to compose sentences and paragraphs in my head when I am dictating as opposed to typing, perhaps because the internal process my brain is closer to coming up with something to say in conversation or a talk. I have always been rather garrulous and have rarely ever been at a loss for words in any situation, so I find it easy to sit here in front of my computer and drone on and on indefinitely.
0 comments:
Post a Comment