Go to previous article
Go to next article
Return to 2005 Table of Contents
Presenter(s)
Paul Hawes
Sensory Software International Ltd
26, Abbey Road, Great Malvern, WR14 3HD, UK
Tel 01684578868
Email: paul@sensorysoftware.com
Overview
Learning outcomes
How to use writing support software in conjunction with text based AAC software to provide users with a route to better text production.
Summary
Many (perhaps most) AAC users have some difficulty with text production due either to slow input speed or problems with spelling. This session will show how the use of software tools designed to help the wider population to write more effectively can also be extremely beneficial in AAC systems.
Abstract
A large number of people using AAC (Augmentative and Alternative Communication) systems have additional difficulties with typing, spelling and word finding. This may be as a result of their condition, or incidental to it. For example, slow writers and typists may take longer to acquire literacy skills, and users with a brain injury may find that spelling is impaired.
As the computer becomes more commonly used as an AAC device, the possibilities to combine other technologies with software that is used for communication increase considerably.
Many of the software tools designed to help speaking people to write can be equally helpful when the purpose of the text is communication.
In this session we will look at programs that provide prediction, word banks, symbol support, homophone-checking, spell-checking and rate enhancement. We will see how they can assist the user to create more accurate text more quickly.
A further possibility that is opened up by the computer platform is a broadening of the meaning of the word "communication". As well as being spoken, any text generated can also be sent as a fax, email or SMS text message. All these methods of communication are very useful to people who use text based AAC, but some users can be shy of using them if they are unsure of the accuracy of their text output.
Introduction
This is in no sense a scientific paper, nor does it represent the results of any academic research. Rather, it is a series of suggestions based on common sense and many years of practice.
I have always taken a broad view of AAC, and I consider that to communicate via email, fax, text messaging or written communication is as important as speech output.
The term "dyslexia" is used here loosely (and quite incorrectly) to cover all manner of cognitive and perceptual problems that interfere with reading and writing.
The illustrated examples are drawn from our own range of literacy support software, but many of the features are also available in programs from other sources. A product comparison is outside the scope of this paper.
Reasons for dyslexia-like problems
There are two main reasons why an AAC user may be especially likely to experience such problems.
1. The disability has an inherent effect on language capability The first case would include strokes and head injuries, or neurological conditions that affect the memory.
2. The user has acquired literacy at a much lower rate as a result of a severe disability The second arises from the fact that even the brightest children learn more slowly if they cannot speak, write or raise a hand when they failed to grasp something.
In both cases, we often find people who have some literacy skills, and a degree of written language, but whose word-finding abilities or command of spelling prevents them from using a text based communication aid very effectively. Many users are worried about using such a system if it will reveal their difficulties in this area, and some people's spelling can be so bad that the speech output is not easily intelligible.
Tools used to assist in text creation
It is increasingly common for people with dyslexia to use a computer with special software to help in the production of correctly spelled text. This type of software will commonly provide all or some of the tools listed below. We will look at each one in turn and consider its use in AAC.
1. Prediction
Predictive typing, or "smart typing" can be used to speed up the rate at which the user can type, but it is also very valuable in helping to get words (especially long ones) typed correctly. The term "prediction" strictly means guessing the word that will follow the one that you have just typed. However, it is commonly applied to systems that offer suggestions to complete the word typed so far. Most prediction systems do both of these. The good ones also learn new words as they are used.
Automatically adding new words as they are used can be a blessing or a curse. In the case of a poor speller, it results in many spurious words being added to the prediction list. Our preferred solution is to add new words automatically, but only when they have been passed through a full spell checker.
Prediction may be used for simple rate enhancement (typing faster) of as a means of spelling support. In the first case, it is best to eliminate any ignored suggestions as quickly as possible. For example, if typing "he" yields a list of suggestions that includes "hello", then typing an "l" would remove "hello" from the list, leaving more room to display other words, such as "helicopter" or "heliotrope".
However, in the latter case, the main clue to the user that an incorrect character has been typed is that the list of words can disappear if an unknown sequence of letters is typed. Therefore, when using prediction to help spelling, all the options should be left displayed.
One optional feature of prediction is the use of a fuzzy lookup. The idea is that the user may begin a word with an incorrect letter (perhaps phonetically selected - like fotograf of nife). The predictor can then suggest words that begin with alternatives such as "ph" or "kn". This can be useful in the production of written text, but is less important for generating speech, as the examples above would be correctly pronounced by any synthesiser.
2. Word Banks
Word banks behave like a prediction dictionary, but with an important difference. Instead of a large selection of general words, the word bank contains only special words that may be useful in the current context.
This is exceptionally helpful to people who can spell everyday words without too much difficulty, but who have trouble with long, difficult or unfamiliar words. A typical prediction dictionary may contain between 5,000 and 10,000 words. However, a list of just a few hundred difficult words related to a particular topic could be very useful indeed.
One advantage of word banks is that, as they are simple lists of words, they can be created very rapidly by a helper. This makes it easy to provide customised lists for use in different situations.
3. Word Bars
Word bars are grids, normally at the bottom of the screen, containing words that may be entered into the text usually by clicking on them, or by selecting with a switch.
Unlike word banks, they are not interactive with the users typing input.
The user may then type common words directly into the system (with or without prediction) and then select difficult words from the word bar.
4. Spell checking and grammar checking
Spell checking is not really so important if you just want to output speech, as most spelling errors tend to be phonetic, and will sound correct when spoken by a speech synthesiser. However, the computer revolution has blurred the distinction between AAC and other computer based activities, so that the same input systems may be used for text production.
As noted above, an automatic spell checker (working behind the scenes) can help to keep the prediction dictionary clean.
Many programs also include a grammar checker (MS Word has a very good one). When the text is to be transmitted later (such as a printed document or an email) many users will wish to use a grammar checker before printing or sending it.
Some programs attempt to incorporate grammar checking into the prediction. This is not especially effective in a language like English where the grammar is so loose. To see what I mean, consider the following sentence fragments:
… the boy
… Eric and the boy
… the policeman looking for Eric and the boy
If the next letter typed was a "w", which words might you predict, and how would you make a rule that decides when to suggest "was" and when to suggest "where"?
Another common problem for dyslexic people is choosing between homophones. These often appear together in prediction or spell checker suggestions. This is unimportant if the output is only to speech, but a problem when communicating in writing. Software that can help the user to decide which word is needed is very helpful. By definition, they will sound the same when previewed with a speech synthesiser, so we need to use alternative descriptions or symbols, as discussed below.
5. Auto replace
This actually covers two very effective tools in one, as auto replace and abbreviation expansion are exactly the same process, although used in a different way.
Auto replacement is used to correct common typing errors as soon as they are typed. For example "teh" could be turned into "the" as soon as the space bar is pressed, and common errors, such as transposing "i" and "e" can be dealt with in this way.
Many text based AAC systems use abbreviation expansion as a handy way to store common phrases. A short word or a simple code is typed, and the auto replace expands it into a whole phrase. However, if the user consistently has difficulty with a word that is needed often then an abbreviation can be used instead.
6. Speech
Speech can be used as a preview before selecting a suggestion such as:
" Prediction
" Word banks
" Word bars
" Spelling suggestions
Where the suggestions are presented in a word bar or grid, the contents of the cell can be previewed by using a right mouse click, or simply by pointing at the cell, depending on the software.
Where suggestions appear in a list, the arrow keys or a mouse click may be used to preview them, depending on the software.
If the list of words includes homophones, alternative descriptions, such as "right handed or correct" and "rite, or ceremony" can help the user to select the correct word.
Another use for speech is in reading text that someone else has typed, or in navigating around menus and dialogue boxes. I have suggested that emails and SMS messages are as integral a part of AAC as speech output. This raises the need for some people to get help with reading the reply. Simple screen readers can give spoken text output from most software.
(Note: For a blind user, reading "most" software is not good enough. Do not confuse the simple screen readers we are discussing here with the far more advanced (and much more costly) programs used by blind people.)
7. Symbols
Symbol communication is often seen as an alternative to text, but in reality many users mix the two. Software that allows symbols to be displayed in a word bar grid can be very useful.
Symbols can also be a good way to help choose between homophones. Symbols are commonly shown in a grid, but some software will also allow symbols to be displayed in a prediction or spell checker list.
Another point to bear in mind is that symbol grids (especially the dynamic screen variety) are very good at getting to a large number of words with a small number of selections. Their weakest point is in handling common function words. For a user who has a good command of common words, an extended dynamic vocabulary may be used to retrieve less common or more difficult words.
8. Stored messages
Most AAC systems allow messages to be stored, normally as a way of reproducing them more quickly. However, it is possible to fill the message lists with messages that contain difficult words, rather than those that are used most frequently.
Summary
Most of the tools used for supporting writing are also valuable to people without speech. I have found a selection of such programs to be as important a part of my assessment kit as my range of switches and alternative voices. I would recommend any speech and language therapist who works with users with emerging or partial literacy to experiment with such programs.
Go to previous article
Go to next article
Return to 2005 Table of Contents