Japanese Version ...................................................... BACK

Intended System


The five systems reviewed do not seem to incorporate distant learning or e-learning. The system I'd like to develop is a server-side one to be accessed for FREE on the Internet and mainly by aphasic adults at home with a help from STs/volunteers at a distance. In other words, it would combine Lingraphica's image dictionary with the web availability of PICOT on WEB and AIST's Pictorial Authoring Tool. Since a nice collection of pictograms has already been provided on the Internet for free, the image dictionary is expected to be developed fairly easily.

What is crucially lacking in the five systems and I believe is strongly needed includes a better representation of sentence structure and a function to make grammatical judgments. Words must be put in a row for us to pronounce simply because we cannot pronounce two words simultaneously! Words, however, depend on one another non-linearly. For example, in the sentence "The man with a mustache likes the lady," mustache and likes are linearly adjacent, but there is no direct syntactic/semantic relation between them. The most important dependency relations are between the verb likes and its subject/objects, and within the subject/object noun phrases, their constituents are dependent on one another. A common way to represent syntactic/semantic dependencies in linguistics is to use phrase structure trees with nodes like NP and VP, but they are not appropriate to be incorporated in an aphasic rehab system.

Exercises on verbs typically found in aphasic rehab books put a verb in the middle with several phrases connected with and surrounding it. Such a representaion of syntactic structure corresponds to a dependency graph proposed in various theoretical frameworks such as Dependency Grammar, Word Grammar, and Link Grammar. Daniel D. K. Sleator, advocating Link Grammar, publicized a Link Parser with its dependeny-based dictionary. The rehab system I have in mind does not have to parse all the possible sentences but only core grammatical constructions of the target language with limited vocabulary. I guess it could be implemented by replacing the pronunciation/spelling of each word in Sleator's lexicon with the corresponding pictogram while inheriting its syntactic dependecy information as it is.

One important modification I'd like to introduce into existing dependency graphs is to make them order-free. As is well-known in linguistics, Joseph Greenberg identifies three major word orders: SVO (e.g. English), SOV (e.e. Japanese), and VSO (E.g. Chamorro). Their word orderings are different, but the dependencies among V, S, and O are essentially the same. I've proposed to use an order-free representation of syntactic structure and derive the three major orderings in terms of the well-known tree traversal algorithms: inorder, postorder, and preorder traversals in my papers. The dependency graphs proposed so far lack this typological consideration.

If a syntactic representaion of a sentence is dissociated from its pronunciation or word order, the representation itself can be universal, shared by any language. Moreover, since the main dictionary of rehab aids consists of images, and images for content words like nouns and verbs are largely universal, the substantial part of the system at stake can be developed by native speakers of any language. I myself would like to work out theoretial aspects of order-free dependency graphs as well as the treatment of function words, which differ among languages.

As for pictograms, a nice collection of them has already been generously offered for free on the Internet. I guess to make them into animation GIFs takes time but is fairly easy. Though I'm not good at such graphical work, I will make rehab exercises with HotPotatoes and free pictograms, which I believe would immediately help people with aphasia. Some of them are found here.

If you are interested in this potentially world-wide project, especially the implementation of the system, please contact me at myasui@dokkyo.ac.jp.

Japanese Version ..................................................................... BACK