What Happens When Students Use Text-to-Speech and Word Predict Programs to Compose Text?
By: Diane DeMott Painter
Two fifth grade students with learning disabilities find writing success using computer technology
Since obtaining my first computer in the early 1980's to use with my students with learning disabilities, I have been fascinated by what seems to be instantaneous improvements in the willingness of most of my students to write. When asked what they think of writing with a computer, most of them respond that it is easier for a variety of reasons. First, they don't have to struggle with forming letters; all they have to do is press the keys and evenly-spaced, perfectly formed letters magically appear on the screen. The text is also formatted automatically with margins. Second, they don't have to "rewrite" the whole composition when revising and editing, they can make the necessary changes and print out their text again. Finally, for most of my students, correcting misspellings is a lot easier, too. Using a spellchecker enables them to spot the errors and make quick corrections by choosing the correctly spelled words from a list of suggested words appearing on the screen.
However, for some of my students, it appears that they have difficulty with effectively using word processing. For some who have poor visual memory, keyboarding is difficult. Even after completing structured and individualized typing drills, they forget where to find the keys. When they do find the keys, they can't visualize how to spell the words they want to write; or they type all the letters in a word, but the letters are out of order. For those with visual discrimination difficulties, when they use the spellchecker, they become confused. So many of the words listed in the spellcheck box look the same. They often complain and point to the words listed, "Which is it check or cheek? " Or, "Which do I use: there, their, they're?"
Students with difficulties in auditory processing deficits often exhibit difficulties hearing differences in letter sounds, making it difficult to apply letter-sound associations when spelling. Their misspellings are so bizarre that the spellchecker often does not have any words to suggest. They will complain, "NO SUGGESTIONS? This computer is dumb!" Finally, their poor auditory memory makes it difficult for them to recall what they want to say and to organize and hold their thoughts in their minds as they type. As a result, sentences may be incomplete, the text will be disorganized, and important details will be missing.
I felt that my students who have problems with auditory processing and auditory memory, as well as visual processing difficulties may be the very ones who were not making the expected gains in writing. They were having problems even when given the opportunity to use word processing and instructed through a structured phonicsbased reading/writing program. So, during this past school year, I decided to try to find ways that my students could learn to "hear," "feel," as well as "see" how letter sound combinations form words. Since I was working with a structured reading program (The Wilson Reading System) to teach my students the phonetic structure of words, I wanted to see what would happen if I used text-to-speech with this reading program to reinforce basic reading and writing skills. This summary of my teacher inquiry illustrates the procedures that I used with two of my students and highlights their progress and my impressions of using text-to-speech and word predict programs with these learning disabled students.
For this inquiry, I chose to study two students who have nearly identical learning difficulty profiles. According to psychological and educational test assessments, both students have auditory processing, auditory memory, visual processing and visual-motor integration difficulties. The visual-motor integration difficulties are particularly evident when they try to "visualize" how letters look and then struggle with forming those letters on paper. As one of the students once explained, "My hand just doesn't do what my brain sees!"
With one student (Abe), I used a text-to-speech program called Talking Text. Writer (TTW) on an Apple II-e computer. This program used a speech synthesizer with a robotic voice to say each word, sentence or phrase as the student typed. Abe had to use his decoding skills to sound out words to type them, or he had to rely on his visual memory of words in order to write. The TTW program pronounced the words as they were spelled correctly or if the sequence of sounds were phonetically written to approximate the correct pronunciation (i. e. lite=light). If a word did not "sound" correct to Abe, he had to either sound out the word and make corrections in the letter-sound associations until he heard what he intended, or he had to use available resources such as a handheld spellcheck device (the program did not have a spellcheck option), a dictionary, or a book. Abe could also refer to his writing plan if he was composing a story.
The second student (Miles) used a different text-to-speech program, Intelli along with a word predict program, Co-Writer. This program also used a robotic voice. It would pronounce each word, sentence or phrase as Miles composed text. The Co-Writer program occupied an area of the computer screen and operated in conjunction with IntelliTalk to predict words that the student may be intending to use in his text. The program predicted words based on subject-verb agreements, word relationships, grammar rules and the frequency of the words that Miles was using in his text. As the predicted words appeared in list form with a number next to each word, the program scanned and said each listed word. In this way, Miles could hear the words pronounced as he studied them. When Miles saw and heard a word he wanted to use in his text, all he had to do was press the word's number. Miles also had the other options to determine how to write words. He could "sound out" words and type them as he heard them pronounced by IntelliTalk, write words from memory, or use available text resources or a handheld spellchecker to find the correct spelling of words.
In the beginning
When I began the study in September, I gave the two students an informal spelling test, the Morrison-McCall Spelling Inventory (List 1) test, to obtain a baseline assessment of their sight word spelling abilities. Both boys spelled nine of the fifty words correctly, indicating a 2.3 grade level in spelling. Using the Wilson Reading Program Phonics assessment test of 38 phoneme sounds covering levels one through four of the program, Abe correctly identified 25 items and Miles identified 24 of the items. These items addressed all single consonants and vowel sound combinations, blends and digraphs. When orally presented with one syllable consonant-vowel-consonant (c-v-c) patterned words to write, Miles generally identified the initial consonant sounds but not the middle vowel sounds or final consonant sounds. Abe generally identified the correct initial and final consonant sounds, but not the correct short vowel sounds. Both boys had difficulty identifying digraphs and consonant blends.
The initial interventions that both boys received were daily Wilson Reading lessons focusing on the consonant and short vowel sounds that make c-v-c words. I would visually present letter cards on the chalkboard and say the individual sounds the letters make and orally blend the sounds together to form the words as I ran my fingers over the letter cards (auditory to visual pairing). To help them recall the short vowel sounds, picture cues were given under the vowel cards such as showing an & for the short "a" sound. The students engaged in a variety of word building activities using these cards, arranging the letter cards as they said the sounds and then blending sounds into words. We focused on "hearing" the sounds as they "visualized" the letters and "feeling" how the sounds are made with the position of the tongue and mouth. Next, both boys were orally given words to "sound out" as they typed (Abe using TTW and Miles using IntelliTalk). This gave them practice typing each lettersound to hear how sounds blend together to form words which the robotic voices of the word processors pronounced. When Abe made an error, he generally detected the error and substituted letters until he heard the word spoken correctly. I noticed that Miles often confused the "soft" sounding consonant sounds, such as m with n, v with f. He was still "guessing" with the vowel sounds. At times he would substitute the sounds and then be satisfied that the word was correctly pronounced. When I had Miles look at me as I pronounced the word and he repeated the word, I noticed that he was misarticulating. He wasn't hearing the subtle differences in the sounds because he wasn't saying the words correctly. We spent more time on "feeling" how the sounds are made (such as where to put the tongue and how far to open the mouth).
By the end of October, both boys were generally able to identify initial and final consonant sounds correctly when spelling c-v-c words. Abe exhibited some difficulty with short o and short u sounds, and Miles was still unsure of short vowels i, e, o and u, although he seemed to know the short a sound. However, he was beginning to hear that the words he typed were not correct, and he also determined that the vowels were probably incorrect. As he typed and heard the sounds in the words, he substitute the vowel letters until he found the correct vowels and the words sounded correct.
By the middle of November, the boys were using their text-to-speech programs to practice writing two syllable c-v-c patterned words with consonant blends and digraphs. Abe was obtaining 80% or greater accuracy when presented orally with the words to type, and Miles was obtaining 60-80% accuracy when presented with the same word patterns. The types of errors Abe generally made involved soft digraphs such as confusing ch with sh. Miles was becoming more accurate with the vowel sounds but he was also having difficulty with discriminating between soft consonants (n for m, and d for t) and digraphs. I noticed this usually occurred when the sounds were in the middle or at the end of words; such as "little" might be written as "liddle," and .1 rabbit as "rabbid."
In addition to the Wilson exercises using the talking word processors, the boys used the computers to type their stories during writer's workshops. Abe readily used available resources such as a novel to type names of people and places for his reading log responses. While writing a reading log response in November, Abe appeared to "know" many basic sight words, such as was and they. He sounded out several other words such as dig and started. He indicated that he knew the word 11 now was not the right word for what he was writing, but he couldn't find the correct combination of letters to make it "new." He also did not appear to hear or see that the s sound was omitted from the word Kings because when he read it back to me after he heard the word processor read it, he read the word with the -s sound. He also did not appear to "hear" or see that the word
The now place was the valley of the king werethey stated to dig. They found some old stone hut.
The new place was the Valley of the Kings where they started to dig. They found some old stone huts.
In November Miles appeared to rely heavily on IntelliTalk to "hear" sounds and rarely looked at the word choices presented in the Co-Writer box as they were being scanned and pronounced. Example 2 shows what he had written as a reading response using only IntelliTalk. He was asked to finish the writing prompt, "Babysitting can be a hard job&133;"
beckos you must no wote do in case of a mrice. You sude no wote to fede him or hir.
because you must know what to do in case of an emergency. You should know what to feed him or her.
The word "mrice" was intended to be "emergency." When I asked Miles to say the word, he said, "Mer-gin-see." Again, this was a good example of Miles writing words that way he pronounces them.
When I reminded Miles that the. Co-Writer box was available to use as a resource for finding words, Miles began to refer to it as he wrote the next sentence for his response. (See Example 3).
You must no hol hot the bottle is so it doesn't bmethe babys throat.
You must know how hot the bottle is so it doesn't bum the baby's throat.
Miles appeared to use the Co-Writer box choices for the words bottle. doesn't, baby (he added the -s) and throat. As he sounded out the words, he typed must, and hol. Because he was saying an -1 sound at the end of the word "how," he typed in the letter -1. He also appeared to have typed several other words from sight memory.
By the middle of December, both boys were showing improved spelling accuracy in the Wilson word exercises, even when presented with two short vowel syllable words containing digraphs and three letter consonant blends. They were also using more sophisticated words in their reading responses.
It also seemed that Miles was relying on the IntelliTalk, feature more than the Co-Writer since he had to be prompted to give Co-Writer a try. This may have been due to our using IntelliTalk with him during the Wilson exercises. However, Miles seemed to notice that when he used the Co-Writer, his spelling accuracy improved because the words sounded as he had intended them to sound.
Abe was also beginning to type text with perfect spelling. He appeared to use words from reading prompts or other available resources. He also was sounding out or using words from sight memory to write the rest of the text.
By mid-year I felt that what I was seeing in their daily writing was significant. Both boys demonstrated willingness to attempt to use more sophisticated words in their writings. Also, the length of their writings was increasing. This was quite evident by the end of January when they drafted their first "story" with little assistance from me. Abe seemed more aware of sound-symbol relationships.
I noted in my observation log, "He appears to be getting the short vowel patterns and putting in long vowel sounds when he hears them. I noticed Abe repeating the words he wants to spell over and over again. It is encouraging to see him becoming an independent writer."
Miles's writing also appeared to be more detailed and he was attempting to use more sophisticated words than before. This meant to me that he seemed more willing to take chances on trying to spell more difficult words. Even though he was still demonstrating short vowel confusions and having trouble with blends, the positions of the single consonant sounds were generally correct.
During most of my observations of his writing, Miles appeared to continue to rely heavily on IntelliTalk to write; although occasionally he did refer to the Co-Writer. However, he was having some difficulty with word discrimination because when he chose the word "Olympics," he thought the robotic voice had said "Olympus."
In the middle of the year
After the winter Christmas holiday I began to introduce long vowel sound patterns "ild", "old," and -e used at the end of a word. It was around this time that it appeared that Miles was considering each letter when entering words into thecomputer. He was sounding out the letters before he typed them. Could it be that Miles was internalizing the sounds? Did he know what sounds he should be hearing and he was using the computer to verify his letter choices?
Miles was also beginning to use more than one syllable words in his writing. The problem was that IntelliTalk did not always pronounce the multi-syllable words correctly. I found that when he "separated" the syllables, the words sounded correct. The other problem was that the longer sounding words were confusing to Miles because they didn't sound as "clearly spoken" as the shorter words. I found this to be true with all three programs, Talking Text Writer, IntelliTalk and Co- Writer. It seemed that he relied almost completely on the computer's voice for the way he spelled words. If it didn't sound right, he would change letters. Sometimes a word was pronounced correctly, but it was misspelled. For example, he wrote shape liss. When I listened to the computer's pronunciation, it sounded to me just like shapeless (no vowel sound difference).
As a result of the unclear pronunciation of the longer words, both boys wrote the words in syllables in order to check each syllable. Then they put the syllables together to form the words. They may have been spelling the words in this manner because in the Wilson lessons we "build" words by syllables with sound cards. They seemed to sense that it was helpful to focus on the word parts in order to get the whole word spellings correct.
I was beginning to be concerned that so much concentration was spent on spelling words correctly that the boys' abilities to compose their thoughts might be hampered. However, the writings of both boys showed complete, well-sequenced thoughts. Both were beginning to show imagination, humor and a real sense of adventure in their stories; so it did not appear that their ability to write expressively was being hampered by trying to spell correctly.
At the end of the year
By the beginning of the last quarter of the year, Abe did not seem to be "studying" each word the computer pronounced before typing in the next word. He was composing at a steady rate with few common word misspellings. In several cases he seemed to ignore what the computer was saying because he appeared to be confident about how to spell the words. He did have trouble with the word "heraldry." He wrote it as herl and he seemed puzzled about the long i sound in the second syllable. When trying to write the word Charlemagne he used in his text to find how it was spelled.
Miles was also composing more quickly without stopping to study each word that he typed. His spelling improvement was not as evident as Abe's; however, he was writing independently. I noted what appeared to be the types of choices he was making to type his words: 1) looking at the Co-Writer box and listening to the words as they were scanned and then choosing the number beside a word that he wanted to insert into his text; 2) sounding out a word as he typed it; and 3) using an available resource to located a correct spelling, such as words found in his writing plan. While writing a 29 word, three sentence beginning to a story, Miles appeared to use CoWriter to spell 21 of the words. He sounded out three words and referred to his writing plan three times to find a word that he wanted to spell. Two of the words he seemed to know by sight. It appeared then that Miles's first choice of available spelling resources was Co-Writer, followed by sounding out words on his own and then checking his spelling accuracy by listening to the way IntelliTalk pronounced the words.
At times, however, I found that when using Co-Writer, Miles chose two words that were similarly configured to the word choices that he intended to use. He didn't realize that the words he had chosen were incorrect, even when he heard the words spoken by the Co-Writer during scanning and when IntelliTalk read the text back to him. The words that he had confused were "than" and "then" and "had" and "head." After this particular writing session, I typed the four words and listened to them pronounced by the robotic Co-Writer and IntelliTalk voices. I detected subtle differences between the words. I concluded that Miles may have had difficulty hearing those subtle differences, and coupled with his visual discrimination and visual memory problems, choosing the correct word from a list of words that all look similar and sound similar may be confusing for Miles.
It was also at this time that we had completed level four of the Wilson word patterns. This level focused on learning to read and spell two and three syllable short vowel words, and long vowels made by open syllables and -e at the end of syllables. Abe was demonstrating 90 to 100% accuracy on the weekly Wilson word assessments, and Miles was making between 70 to 80% accuracy. These weekly assessments were spelling pretests and posttests of word patterns presented during the Wilson lessons. However, only Abe was fairly consistent with applying those patterns in his daily writing. Miles continued to demonstrate difficulty with misarticulations (repeating the sounds and words incorrectly) and these mispronunciations were reflected in the way he spelled words, especially soft consonant sounds such as confusing m for n, d for t. He also continued to demonstrate vowel confusions. At times, when he tried to write short vowel patterns but he couldn't identify the correct vowel, he would leave out the vowel from the word.
The Morrison-McCall spelling inventories given every two months also showed spelling gains throughout the year, although it appears the most dramatic gains were made from February to April.
|Morrison McCall List||List1 Sept||List2 Nov||List3 Feb||List4 Apr|
Table 2 Spelling Gains
I also administered the informal Wilson sounds posttest to determine any improvements in the boys' abilities to identify phonetic patterns when they were visually presented with sound cards. Both boys showed improvements. I considered 80% accuracy to be a mastery level score. Abe reached mastery with his score at 82%, and Miles was close to mastery with a score of 76%.
1 had also noticed that their oral reading skills were improving. They were using both context clues as well as sounding out multi-syllable words when they did not know a word. In addition, after reading a story selection, both boys appeared eager to use their computers to make reading responses instead of making hand written entries -in their reading logs. They also began "checking on" each other, sharing their written responses and offering comments and suggestions or explanations about events related to what they had written.
When I chose the text-to-speech processors (Talking Text Writer with Abe and IntelliTalk with Miles) I had three assumptions: 1) text-to-speech will provide auditory feedback which will either confirm or invalidate the correct sounds the children think occur in words; 2) seeing and hearing the sounds that text-to-speech programs produce will provide the students with multi-sensory practice in learning how to write words; and 3) as the students learn to distinguish sounds and think about soundsequence through text-to-speech programs, they will begin to internalize how words are formed each time and will be better able to recall words from memory instead of "sounding out" words and relying on hearing the words spoken back to them as they write.
I had two assumptions when I chose the word predict program Co-Writer to use with IntelliTalk.: 1) Co-Writer would be useful in the sense that based on the context of the writing, the program would provide Miles with a list of correctly spelled words that he could easily insert into his text; and 2) that by setting the program to scan and saying each word in the list, he would be able to see as well as hear each word (a multi-sensory presentation) which would enable Miles to choose a correctly spelled word.
For the most part, both the text-to-speech programs, Talking Text Writer and IntelliTalk , provided the students with the opportunity to see and hear the sounds that make up the spelling of the words they were writing, and therefore helped the students think about correct sound sequence. When the students typed one syllable c-v-c patterned words, they were generally able to determine if they had spelled the words correctly, and when they made a spelling error, they were generally able to "hear" the error and make a correction. However, when typing two or more -syllable words, the pronunciations of the robotic speech voice were sometimes unclear, making it hard to clearly hear all parts of multisyllable words.
It also appeared that as the children learned to distinguish sounds and think about sound-sequence through text-to-speech programs, they began to internalize how words are formed and were better able to recall words from memory. It appeared that both boys demonstrated less "sounding out" behaviors of the words that they typed. It seemed that by April they were becoming more confident with their spelling and they were using words from memory. They still used the "read back" features of the programs, however. It seemed that they enjoyed hearing what they had written.
The predictability feature of Co-Writer was also fairly impressive in the sense that it usually produced a list of words that Miles needed in his text. The scanning feature was also useful in the sense that it highlighted and pronounced each word in the predicted list. I was able to control the speed of the scan and the rate of speech which I also found to be a useful feature. Most often Miles was able to correctly identify a word that he needed. However, there were times when Miles had difficulty identifying a correct word because of his visual discrimination and auditory discrimination difficulties. Similarly configured words and unclear pronunciations made it difficult for him to correctly choose multisyllable words, even when the different robotic choices of IntelliTalk were used. Words such as "Olympus" and "Olympics" looked and sounded alike to him.
Suggestions for future program development
The development of text-to-speech programs and word predict programs should be made with the considerations of the learning strengths and weaknesses of the users. My students with their auditory discrimination problems found the different robotic voices hard to understand at times; particularly long words. Perhaps as technology advances and text-to-speech programs move from using robotic voices to using more natural speech pronunciations text will be better understood. It would also be more effective if similarly configured words were not listed one on top of another, but were listed separately. Words with the same consonant letter sequences but different vowels seemed confusing to Miles who has visual discrimination problems. When presenting the predicted word lists, perhaps the words should be shown inlarge print and placed across the bottom of the page, separated by boxes. In this way, the program can highlight each box from left to right as it reads the words and the words will be easier to recognize.
In conclusion, I think developing technology, such as word processing programs that are multi-sensory and easy to use, should remain a focus of both hardware and software development. As newer and better technologies are developed with differences in learning styles and functional needs in mind, the greater the chance that special needs populations will gain independence as writers and learners and demonstrate success in mainstream environments.
Co-Writer. Don Johnston Developmental Equipment, Inc.
P.O. Box 639, 1000 N. Rand Rd.
Wauconda, IL 60084
5221 Central Ave.
Richmond, CA 94804
Morrison-McCall Spelling Inventory.
Spaulding Education Foundation
2301 West Dunlap Ave., Suite 105
Phoenix, AZ 85021 Tel: 602-266-9158
Wilson Language Training.
162 West Main St.
Millbury, Mass. 01527
Diane DeMott Painter (2002)