The Living Schoolbook

When I wet to the Closing the Gap conference in 2003, I was really amazed by what teachers were doing with software that I already knew.  I had used each of these pieces of software, but the combination of the two in this manner really caught me off guard.  When I came back to Syracuse I told quite a few people about this.  In fact, I still am telling people how cool this was.  Here is what I wrote. 

Voice recognition software like Dragon Naturally Speaking has been used to allow people with motor disabilities to write on their computer by dictation and give commands to the computer to control it. Users typically go through a training process where they read text from the screen to "calibrate" the speech recognition software. The longer the user uses the software, the better the recognition becomes and fewer errors are made. In practice, the use of voice recognition software with those with motor and speech disabilities has been difficult and yielded poor results.

Norman Rothstein from the Cabarrus County School District in Concord, NC presented a combination of using speech recognition (Dragon NatuallySpeaking 7) software with word prediction (CoWriter) software. Students with dysarthric speech trained voice recognition software with a "command only" subset of the normal routine. As such, the voice file for that user was a small vocabulary of commands. Norman reports that the smaller vocabulary reduced the error the software had in recognizing distinct commands such as "type" and "press". By using the voice recognition software to have students "type" letters into CoWriter and the select a predicted word the students were writing at 25 wpm.

An example may go like this:

• Student: "Press A", "Press P" (CoWriter shows words starting with "ap". Apple may be the third choice.)
• Student: "Press 3" (CoWriter types "apple")

and so on...

Some attendees mentioned that too many letters sound the same (like c and e or m and n). Norman said that the students could use "alpha" for "a" and "beta" for "b", etc. This seemed to fix the problem.

He also demonstrated how the "mousegrid" program that comes with NaturasllySpeaking could be used to control the computer:

• Student: "mousegrid" (a numbered grid appears on the screen)
• Student: "Press 1" (a new grid in the 1st square of the original grid appears)
• Student: "Press 5" (a press of the button on the screen in square 5 occurs)

Norman also says since the student only trained the computer on certain commands, background noise and comments not related to the computer task are usually ignored by the voice recognition software. This has allowed students to use the system in the classroom effectively.

I think this was the coolest thing I saw at the conference.