DYSLEXIA
TEACHER

Voice Recognition Software Helping Dyslexics

by Ian Austen

ANDREW GANAT graduated from high school in May and is looking forward to college in the fall. While that is an important milestone in anyone's life, reaching it is especially sweet for Mr. Ganat, 18. In the second grade, he was found to have the learning disability dyslexia.

"My overall problem is memorizing the spelling of words," said Mr. Ganat, who is from Norwalk, Conn. "I can get all the letters down but I can't make them into the word."

Specialized schooling — including five years as a boarder at the Gow School near Buffalo, which specializes in teaching students with language-based learning problems — certainly helped take Mr. Ganat to the threshold of college. But for the last three years he has found another way to boost both his grades and his outlook. Mr. Ganat is one of a number of dyslexics, both students and adults, using voice recognition software to transfer their ideas into print.

"It's made my grades improve," Mr. Ganat said of the software. "I didn't think I was getting the right grades before. I'm not a dumb kid, so it was frustrating. But now I've kind of found a way to be even with everyone else."

Marshall H. Raskind, a learning disabilities researcher at the Frostig Center in Pasadena, Calif., said that voice recognition software could make a significant difference for many people with dyslexia. "It is a great equalizer," he said. "When someone feels they can express themselves in writing it can have positive implications for self- esteem. One guy told me, `For the first time in my life I can write love letters.' "

After studying the use of the software by dyslexic students for 10 years and publishing four joint papers on his findings, Dr. Raskind has concluded that speech recognition not only allows dyslexics to communicate more efficiently but may even help them overcome their condition.

"Children who wrote using speech recognition technology for as little as 10 1/2 hours showed significant improvement in reading, decoding, spelling and comprehension," Dr. Raskind said. "We were blown away by this. The results are preliminary. But it is very encouraging."

The use of voice recognition software by dyslexic students has largely taken software companies by surprise. "The focus is around driving larger business opportunities for voice recognition," said David Nahamoo, senior manager for human language technologies at IBM. (news/quote) Research. "But the last time I saw the letters from dyslexic users, I thought, it's wonderful."

While software companies have focused on developing voice recognition for common uses like controlling cell phones, making computers more accessible to non-typists and hands- free control of gadgets in automobiles, the technology is slowly making its way into a range of applications for people with disabilities.

Harnessed by determined researchers in what is almost an underground movement, it has helped people with impairments ranging from paralysis to repetitive stress injury that make typing painful or impossible. People with dyslexia are now beneficiaries, too.

Dyslexia is a broad term for language disabilities that cause a person to have trouble understanding written words, sentences or paragraphs. The International Dyslexia Association estimates that dyslexia is the most common source of reading, writing and spelling problems. Dyslexic students tend to have separate vocabularies for writing and speaking: even if they are highly articulate, they draw on a strictly limited selection of words when writing. When they were able to dictate their papers and examination responses to a computer, Dr. Raskind found, the students exploited their full language capabilities.

So far, the use of voice recognition by dyslexics is very limited. Dr. Raskind said he had been unprepared for the opposition he encountered from some people, including teachers.

"I don't want to make it sound like a panacea," he said of the software. "It can be very, very frustrating for some students. But many people view assistive technologies in general as a crutch, a way of avoiding a problem. It's weird: it's like seeing someone with a white cane and saying, `Rip that cane out of their hands and let them do it themselves.' "

Robert Follansbee, co-director of Speaking to Write, a project financed by the Department of Education that is examining the use of voice recognition in secondary schools, said: "Special educators are hip to it now. They get it. But often regular educators don't understand it. The comment I've heard many times from teachers is `They'll never learn to write.' "

Dr. Raskind first got the idea of having dyslexics rely on computers when he headed the learning disabilities program at California State University at Northridge. In 1991 he began developing a long-term research program with Eleanor L. Higgins, a senior research associate at the Frostig Center, a research and treatment center for people with learning disabilities. Together they focused on dyslexic California State students who were dictating reports and exam answers by using Dragon Dictate, a precursor to the popular L&H Dragon Naturally Speaking software.

"Often they'll be able to talk it out fine but they have difficulty translating it to the printed page," Dr. Raskind said.

Dragon Dictate's performance with what was then state-of-the-art computers was, Dr. Raskind recalls, not brilliant in technical terms. But the results generated by the 29 Northridge students who used it were impressive. "You could no longer differentiate their writing when they used speech recognition from writing by students without learning difficulties," he said. "The quality of their writing was far superior" to what it had been.

So were the students' marks. One factor was what might be called the 25-cent word effect. "If you use bigger words, a bigger vocabulary, you get graded higher," Dr. Raskind said.

It was some of Dr. Raskind's own subjects who first suggested that the software might even be having an effect on their reading or writing skills. "After using it over the course of a year, they started saying things like, `You know, I think my reading and spelling are getting better,' " he said. Two joint studies he has done since then tracking students ages 9 to 18 seem to confirm their impressions.

Dr. Raskind believes that the explanation may be fairly straightforward. "You say a word and then you see a word," he said of the programs. "That's an age-old approach that's used with kids who have dyslexia."

Mr. Ganat, the college-bound 18- year-old, began working on voice recognition in the 10th grade. Although he struggled at first, his main problem now, as he tells it, is dealing with others' disbelief. Using the IBM. speech recognition software Via Voice, he writes the first draft of most of his papers and even the answers to some examinations. Then he pastes his work into Microsoft (news/quote) Word for revisions and corrections.

This year he wrote a 15-page history paper on the construction of the Pentagon. "My history teacher didn't believe I wrote it at first because it had no spelling mistakes," said Mr. Ganat, who plans to attend Johnson & Wales University in Providence, R.I. "Now I feel like I'm ready to write at a college level."

Since switching to the software, Mr. Ganat said, his reading has improved "but my spelling is not up to par."

Certainly Mr. Ganat's mother, Elaine, has noticed a striking improvement in the e-mail he has sent home. "Before you had to figure out where the vowels went," she said. "Now you can read them."

Brett Jeremy didn't have the advantage of voice recognition in dealing with his dyslexia in school. He got through college by painstakingly writing and rewriting papers in longhand and then turning them over to a girlfriend "who could word process at 80 words per minute." After becoming vice president for production at Native Kjalii Foods, a maker of salsa and tortilla chips in San Francisco, it took "days of frustration and pressure" to produce a five-page technical document because he constantly transposed letters within words.

About four years ago Mr. Jeremy tried an early version of Dragon. "I thought it was the stupidest thing in the world," he said. "It took forever and I didn't understand it." But he gave a subsequent version of the software a chance a year later and became a believer. "Now I can push out a very technical five-page document in less than a day," he said.

That's not to say the program is perfect. "Try dictating a word like `dewatering granulator,' " he said. "The software just doesn't even try."

Paradoxically, some of the technological advances being made as IBM. and L&H chase a wider market may work against the interests of dyslexics. Early voice recognition programs demanded that users utter each word slowly and deliberately. The results, after a bit of delay, appeared on computer screens one word at a time. Current software, however, demands that users speak at a normal, conversational pace, with the words quickly flooding the screen.

"It depends on the student, but for some it can be too much information, too fast," Dr. Raskind said. He is one of many researchers who think that the answer may be a return to the past and are studying the features of voice recognition software to find which are most effective for dyslexic users.

The goal is to develop a stripped- down version of the software that won't overwhelm the user — a feeling that even the fully literate consumer knows only too well.

With thanks to the excellent and informative New York Times (July 19th 2001).

• Computing Out Loud - by Susan Fulton
• Dragon Dictate
• IBM Via Voice