2004 Conference Proceedings

Go to previous article 
Go to next article 
Return to 2004 Table of Contents 


Russell Thomas Cross B.Sc.(Hons.), DipCST, MRCSLT.
Prentke Romich Company
1022 Heyl Road, Wooster, OH 44691
United States

The Prentke Romich Company's SpringBoard portable, touch-screen communication aid, has recorded speech, which makes it language independent. Structured language applications have been developed for a number of European languages. These programs use a word-based approach to provide generative language potential.


The real challenge in developing an AAC system is how to provide a word-based generative system. By combining single words into sentences, an individual can create novel utterances, or true language.

Consider the following sentences and imagine what we need to do in order to build a simple communication board:

"I want something to eat"
"I live in Wooster"
"I like that"
"I feel good"

By taking all the different words in the example sentences, we can build a manual board that will let us generate not only the already identified sentences, but also other new sentences.

sixteen key grid containing sixteen different words

To the 16-location board has been added three extra words; "you," "it," and "is." This simple addition provides a much more powerful vocabulary set. It is now possible to generate new sentences:

"I want this"
"I want to live in Wooster"
"It is good"
"It is good to eat"
"I like to feel good like this"
"Is that something you like to eat in Wooster?"
"I want to live in something that good"

Notice how all of these were NOT in the original set, yet by using a word approach it is now possible to produce such utterances.

The next step is to increase the vocabulary size even more by considering how some of these words change depending on how and where they are used. For example, we can say "I want" but not "it wants," "I want this," but not "I want that." The changes we need to make are therefore to add (a) the "-s" form of the verbs and (b) the word "that" - the "far form" of the determiner, "this." So let's take a look at an updated board:

thirty six key grid containing twenty eight different word

Also added to the set are the words "am" and "are," which are simply forms of the verb "to be" and are wholly predictable from the preceding or proceeding pronoun: If the pronoun is "you/we/they," it must be "are;" if the pronoun is "he/she/it," it has to be "is."

But this is where the challenge comes in. As the inevitable vocabulary size increases, so does the size of the grid. To address this, the usual solution is to split the grid into many smaller grids or pages. For a 3,000 word vocabulary with 30 words on a grid, this means a minimum of 100 grids.


Instead of using single grid locations to represent single words, it is possible to use sequences of grid locations to represent single words. Furthermore, by using pictures instead of words, the benefits of this sequencing can be made available to non-readers. To illustrate this notion, consider the two small grids below, which can be used to generate three pronouns and the "to be" verb:

two grids, one with six different words, the other with four different pictures

Notice that with the first grid, you need six keys to handle all the possible combinations of the words, but in the second, you only need four. Why? Because in the second, the picture of the bees is used for ALL forms of the "to be" verb. The table below shows how the actual phrases are stored:

Two by six grid showing two-hit sequences for phrases

The sequencing also handles the peculiarities of English grammar, specifically the inflection of the "to be" verb.


Using sequencing is also a useful strategy in non-English languages. For example, in French, the word for "I" is "je," but it changes to simple "j" when followed by a verb that begins with a vowel or a "h." So, "I think" is "Je crois," but "I like" is "J'aime," not "*je aime." So only one key is needed to represent "je," not two for "je" and "j". This pattern is used in the French application program.

36-key overlay showing French phrases beginning with I

Similarly, in the German application, phrases beginning with a pronoun can be stored as two-hit sequences with verb changes being handled by the device, rather than the client having to choose a form.

36-key overlay showing German phrases beginning with I

Dealing with verb phrases is just one way in which sequencing can be useful. It can also operate in representing closed-class word groups - those words that constitute a small, identifiable, grammatical class that doesn't change (e.g. determiners, prepositions, pronouns etc.) In the Spanish application, all question words start with an icon showing a question mark:

36-key overlay showing Spanish question words

In figure 7. notice that not only can you have question words but also verb forms. This is another feature of icon sequencing that can be used to benefit the client: The potential to use a specific icon to represent multiple words. Although this sounds intuitively confusing, it actually becomes very powerful and means that large vocabularies can be stored using small icon sets. This in turn decreases the need to have many hundreds of grids or pages to represent large vocabularies.


Coding language in a voice output communication aid is a challenging task. The larger a vocabulary set becomes, the more grids/locations are needed. Using icon sequencing reduces the need for multiple grids. Furthermore, it helps to handle many of the grammatical inflections of different languages in a systematic way.

Go to previous article 
Go to next article 
Return to 2004 Table of Contents 
Return to Table of Proceedings

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