1999 Conference Proceedings

Go to previous article 
Go to next article 
Return to 1999 Conference Table of Contents


USING MANUAL WORD-PREDICTION TECHNOLOGY TO CUE STUDENT'S WRITING: DOES IT REALLY HELP?

Dr. Colin J. Laine
Mr. Tony Bristow
The University of Western Ontario
Information Services Inc.
1137, Western Road
St. John's, Newfoundland, Canada
London, Ontario, Canada N6G 1G7
t.bristow@is-inc.com 
laine@julian.uwo.ca

A continuing concern relating technology designed to assist written expression to students with disabilities, is discerning "what works". Assistance through word prediction has been extensively promoted in the past decade as a productivity tool for all; but, does its use help a writer's productivity? Maybe. Word-prediction is like a coat of many colours - its hue depends on the author or the vendor. Despite vendors' claims of considerable positive results, the results of research into this area are equivocal (cf, Koester & Levine, 1994; Sommers, at al 1994;Treviranus & Norris, 1987; Venkatagiri, 1994).

There are two types of predictive dictionary: oral word-prediction and manual word-prediction. Laine & Wilkinson (1997, under review) concluded that oral word-prediction (speech-to-text) requires a higher cognitive functioning level as well as a good language ability for it to be an effective writing tool. This conclusion led us to focus our studies on manual word-prediction (keystroke-to-text) technology to assist word-finding with low-language children and adults. As each keystroke is made, the window containing the word list is modified to predict the word (or phrase) from the main dictionary. Our work has focussed on elementary and secondary language-impaired students in rural and urban schools with a range of learning environments from full regular class inclusion to separate class settings.

Students with impaired or emergent language skills are hampered by failure to recognize the words they wish to use (Bjaalid, Hoien, & Lundberg, 1995; Bruck, 1993). Their written language is improved significantly through intervention in word finding, word fluency, and contextual cueing (Corrigan & Stevenson, 1994); and, word-prediction (Heinisch & Hecht, 1993; Laine & Follansbee, 1994). If, as Wiig & Semmel (1980) showed, word-finding skills can be assisted considerably through cueing, then how can we effectively assist cueing for low-language learners?

"Look it up in a dictionary...use the spell-check..." is pointless advice unless you already have some idea of what the word is like. The critical task is to get past the wall of words.

Placing a word-list in the hands of the writer so that words or phrases emerge as the letters unfold would cue the writer by predicting what word might be wanted; adding an aural cue would be especially useful to visually dysfunctional writers. For this assistance to work most effectively for low-language populations, we would need to have ready a predictive dictionary that could also be sensitive to context.

Therefore the addition of content word-lists or "lexical-chain" technology has become another important consideration. A package conceived as a writing "toolkit", which would allow users to switch on any of the tools they might need, would provide the most effective, individualised environment. The writer would thus be able to maintain a semantic flow by choosing settings; choosing the word or phrase offered; copying the letter patterns; or using specialised keystrokes. Maintaining a flow of ideas is one of the keys to effective language-use, problem-solving and good written expression.

WriteAway2000(R) (I.S.I. 1998) was designed as a "toolkit" that presents itself on-screen as a typical word-processor.

WriteAway2000 affords individualised access through specialised content specific word lists; student's personalised dictionaries; stylised speech output; and variable environments for access by students with multiple impairments: providing the teacher with a program that will allow for maximum latitude in an inclusionary environment. The user can decide, and make available those tools needed to complete the task; thus a teacher and a student make a conscious decision to personalize the environment and to use specific words in the context of a particular writing project. Such words appear at the top of the "high list"for that project but will not be added to the main dictionary: cluttering it with infrequently-used words.

By contrast, the prediction found in speech-recognition systems is phonemically based. A combination of phonemes dictate which words in the computer's dictionary are "probable fits". Effective use of this type of system requires the user to have knowledge of word structures and patterns. Without such knowledge, the user's voice-files (like voice prints) become corrupted and particular sounds become attached to erroneous spelling (e.g., the oral "dodge" is translated to the letter combination "d-o-g-e"). Low-language users may become frustrated with oral input when readers point out a number of errors which the computer did not understand.

Alternatively high-language, low-speech users would find the speech-recognition systems very effective at understanding them - maybe more so than their colleagues. It is critical therefore to ensure that the cueing device is appropriate to the needs that have been assessed. For a student with very low language facility, a manual prediction system provides more effective cueing than a sophisticated ("sexy") speech-recognition system. Also we have found that such students have a greater sense of control over slower manual word-prediction than they do over speech-recognition systems.

Allocating 'tools' that are appropriate to the students' needs and abilities has resulted in significant increases in written expression by minimal language proficient students. They wrote more when we divided the writing process into separate components (conceptualising, mechanics, and publishing). As the manual word-prediction users wrote more, they gained access to a wider range of words which encouraged a greater flow of ideas. By contrast, the oral-system users showed that while they wrote for longer periods, their flow of ideas were discouraged as they had to focus on accurate word recognition.

We can confirm that using manual word-prediction assisted the students' written expression primarily in Word Finding, Word Fluency, and Word Complexity. We are currently examining the flexibility of WriteAway2000's other tools and how they might provide a ready platform for inclusion. While we have demonstrated the use of word-prediction is associated with improvements in our subjects' writing and productivity, the key element has been providing appropriate training and ongoing support to teachers, who in turn have provided ongoing support to the students.

We also found that the greatest acceleration in written expression has been experienced by those students with the more diverse communication abilities; where the teachers were fully conversant with the program and used it themselves; where the teachers have incorporated the writing tool as "normal" for everyone in the class regardless of whether or not they had been identified as exceptional and in need of specialised technology; and where the teachers encouraged the students to use any tools as they thought appropriate. Thus any of the classes might have twenty or thirty students using WriteAway2000 with as many variations in the preferences (tools) being set, but their not showing on the computer screens nor determining the teachers' responses to the students' compositions.

References

Bjaalid, I-K.; Hoien, T.; Lundberg, I. (1995) A comparison of components in word recognition between dyslexic and normal readers. Scandinavian Journal of Educational Research. Vol 39(1) 51-59.

Bruck, M. (1993). Word recognition and component phonological processing skills of adults with childhood diagnosis of dyslexia. Special Issue: Phonological processes and learning disability. Developmental Review, 13(3) 258-268, Corrigan, R., & Stevenson, C. (1994). Children's causal attributions to states and events described by different classes of verbs. Cognitive Development, 9, 235-256.

Heinisch, B., & Hecht, J. (1993). Predictive word processors: A comparison of six programs. TAM Newsletter, 8, 4-5,8-9. Information Services Inc., (1998). WriteAway2000: The all-in-one toolkit promoting written expression and literacy. St. John's Newfoundland: Author.

Koester, H. H. , & Levine, S. P. (1994). Learning and performance of able-bodied individuals using scanning systems with and without word-prediction. Assistive Technology, 6, 42-53. Laine, C. J., & Follansbee, R. (1994). Using word-prediction technology to improve the writing of low-functioning hearing-impaired students. Child Language Teaching and Therapy, 11, 283-297.

Laine, C. J., & Wilkinson, W. (1997, under review). Improving employment opportunities for persons with disabilities through training: A case study using speech recognition systems. Training & Development.

Sommers, R. K. et al. (1994). Word skills of children normal and impaired in communication skills and measures of language and speech development. Journal of Communication Disorders, 27, 223-240.

Treviranus, j. & Norris, L. (1987). Predictive programs: Writing tools for severely physically disabled students. Proceedings of the 10th RESNA conference. Pp 130-132.

Venkatagiri, H. S. (1994). Effect of window size on rate of communication in a lexical prediction AAC system. Augmentative and Alternative Communication, 10, 105-112.

Wiig, E. H., & Semmel, E. M. (1980). Language assessment and intervention for the learning disabled. Columbus, OH: Merrill.


Go to previous article 
Go to next article 
Return to 1999 Conference Table of Contents 
Return to Table of Proceedings


Reprinted with author(s) permission. Author(s) retain copyright.