What’s your perspective on speech-to-text software?
For some students with writing challenges, the thought is, “Oh my goodness, if you’re having difficulties with spelling, why not just talk to your computer?” The challenge with talking to your computer is the computer literally writes down word for word, and sound for sound on what you say. Now, six years ago, to be able to train a voice recognition program, it took about six hours to train at around an 80% accuracy rate. So if I dictated 100 words, 80% or 80 of those 100 words would be recognized properly.
Today, with absolutely no training at all, we’re at about a 98% accuracy rate. So right out of the box, I can sit down and start talking to the computer, and it will write down 98% of what I say perfectly. The challenge though is, as I’ve been talking here, I haven’t actually said any punctuation. So in spoken language, we don’t say periods, commas, quotation marks, we just speak, and the listener kind of fills in where the gaps are. But often, when you’re talking to the computer, you have to dictate all that information. You have to say, “How are you doing today? Question Mark.” “I’m doing fine. Period.” So you have to dictate your grammar right into the software.
In addition, a lot of students don’t actually know how to speak in a way that translates into formal writing. Formal writing has its own unique style. Oral communication has its own unique style, and being able to talk to a computer in the way that writes out a nice perfectly structured paragraph is very, very challenging for most students. So, though voice recognition may seem like a great technology that’s going to help so many students, what we actually find is the amount of time it takes to edit what a student has dictated to the computer can be even longer than if they just typed it up themselves.