Wednesday, September 26, 2007

Software review: iListen voice recognition program by MacSpeech

I have been using a voice recognition program called iListen for several weeks. Overall, I am satisfied but it can be extremely frustrating at times.

First, the good part: it allows me to effectuate one of the main rules espoused by my legal writing instructor during the first year of law school. That rule is as follows: start writing. Just get something on paper. You see, my first year writing instructor was very similar to me. That is, he was a major procrastinator. And if there is one thing we procrastinators are good at, it is not starting to write when we should. The solution to that ailment is simply to start writing, even if you know nothing about the topic. This will accomplish two things: first, you learn that you did know something about the topic - more than you thought you knew. Second, more importantly, you will have started. And starting is half the battle. Procrastinators - chronic procrastinators - get that way because they become paralyzed by a feeling of dread.And has the deadline approaches, and nothing has yet been written, the dread only increases, as does the paralysis. By getting something on paper, even if you have to, as my instructors so colorfully put it, "vomit eight pages onto the screen," you now have something to work with. You now have a first draft. And it is much easier to revise and start tweaking when you have something already written. Even if it sucks.

Take, for instance, the letter that I wrote to the commissioners of the FCC. For the last two months, since my indecency paper was published, I had been intending to send a copy to each of the five commissioners of the Federal Communications Commission, along with a brief introductory letter explaining what this paper was about. Yes, I intended to write this letter every week. But the blank screen was staring at me, and even though I had just written and gotten published a 15,000 word essay, this simple 200 word letter was eluding me. There's something about hands on a keyboard, with a flashing cursor staring at me, that simply serves to clear my mind, like some sort of Zen-like trance.

The voice recognition software solves that problem. Yes, it makes simple mistakes, because it misunderstands the easiest words. But it gets most of the thought onto the screen. And I have found that speaking something aloud is far more natural and quick than trying to force my fingers to type the same thoughts. (I now see the appeal in dictating memos and letters!)

So, where it had taken two months and I had not written anything, with the help of the voice recognition software I now had a complete letter - personalized for each Commissioner - written in the time span of approximately half an hour. As I speak this entry, five copies of my article are sitting in a mail truck somewhere, and tomorrow they will be sitting on the desk of the commissioner (or at least the commissioner's assistant).

Now for the most frustrating part (other than the constant incorrect words): the vocabulary of this software program is simply not large enough. Friends, my vocabulary is not that sophisticated. I don't think I should have to teach the software what the word sneeze is. Or what all the derivations of the word are - sneeze, sneezed, sneezing, etc. This problem happens so frequently that I continue to spend several minutes during each dictation session teaching the software words that it simply has no excuse for not knowing. The programmers say that if they included all possible words, it would be harder to dictate correctly, because the computer we have so many more choices - so many more words to choose from. But folks, honestly, this is too much. Below is a list of the words that I have had to teach the computer during this blog entry alone:
  • espoused
  • procrastinators
  • colorfully
  • vomit
  • eluding
  • Zen-like (okay, this one I can understand not including in the Standard dictionary)
  • misunderstands
  • aloud
  • sneeze
  • sneezing
Honestly, there are more, but I have forgotten which ones I had to teach the computer. This happens all the time. Obviously, as I continue to use the software more frequently, my particular brand of language will be incorporated into the vocabulary database, and I will have to make these additions less frequently. Until then, it is incredibly frustrating.

And there is the fact that sometimes the program will be almost 100 percent accurate, and sometimes the accuracy will offer somewhere around 60 or 70 percent. The biggest variables that determine how accurate the software is all our position of the microphone, and background noise. Surprisingly, the program is still fairly accurate even when there is light music playing in the background. But move the microphone one inch away from the proper position, and watch out!

And then there are the smaller annoyances (edit: I just had to add the word "annoyances" to the dictionary! how ironic!) -- this software uses a special correct mode in order to fix incorrect transcriptions (edit: I just had to add the word "transcriptions" -- it had the singular but not the plural!). It gives you a list of words to choose from - alternate possibilities - and use select the correct word, and you hit done. Now, the way this software actually makes those corrections is by constantly keeping track of where the cursor is on the screen. However, this method does not work very well. Help files that come with the software remind us not to manually reposition the cursor at any point during the transcription or else problems will occur. However, I have been very careful to not manually move the cursor, and I still see frequent incorrect placements of the cursor when coming out of correct mode - either one space too far off, or one space too far back. The end result in is odd spacing on the final product, such as too many spaces or too few spaces between words or sometimes even inadvertently cutting off some leading or trailing characters.Simply put, you have to be very careful and attentive when using the transcription software. It is tempting to simply close your eyes and lean back and let out a stream of consciousness. But you may not be pleasantly surprised at the end of your beautiful stream.

Even with all of those caveats, I have found the voice transcription software very useful. I may just have to take the precaution that one of my communications law professors took: include as part of my e-mail signature a line that sense, "please excuse any odd errors caused by voice recognition software."

Now, if you will excuse me, I have been sitting here talking long enough. I am sorry for the inordinately long post, but voice recognition will tend to do that!