-
Notifications
You must be signed in to change notification settings - Fork 1
Thoughts Almost a roadmap
I am thinking that there should be a json file (possibly output from a database, and possibly several files) to define the tiles. Each one would have fields for:
- title
- alternative pronunciation text
- other language text (in fact a common language structure would be a good thing to put in from the start)
- combining rules (what should happen to it when said next to other words e.g. should it have "a" or "an" in front of it)
- image url
- x position (number of tiles from the left)
- y position (number of tiles from the top)
- array of sub-tiles
- link to another json file of sub-tiles
Separating the tile data like this would allow people to create their own personal tile sets (I can imagine a "my family" tile set with photos for the icons) combined with community edited tile sets. Possibly a database driven back end for managing tiles would be plugged in here.
it would be great to have a means of combining packs of tiles, possibly some non-Free ones using licensed icon sets. The main focus will be on creative commons licensed re-distributable artwork, but there are some specific AAC icon sets which some people may already be familiar with.
the tile background colour could mean something useful. Maybe separate things and actions (verbs and nouns).
If you press "want" and "biscuit" then hit a button to say it nicely it should say something like "Please can I have a biscuit?" by arranging the subject and action into a grammatically correct sentence and figuring out if it should be a question etc.
Maybe a couple of modifier buttons would be good to express the concept of "big/more" "little/less" so you can say "lots of rain" or "a little bit of snow" or "big brother" or "big glass of juice with a little ice" they could also apply adaptively to verbs selecting an appropriate adverb, e.g. "run quickly". "big" + "want" could mean you strongly desire something.
Right now you press a button and hear it. It would be good to add the button concept to a queue and then say a sequence of buttons at once - possibly reinterpreting what is said to express a sequence of concepts in a full sentence.
There are voices optimised for different languages. Tiles could be translated so that people can converse in other languages. This might go with specific tilesets for certain use-cases such as travel (vocabulary for catching trains/planes/booking hotels etc.)
It might be a good idea or a terrible idea to pre-load all the audio clips from the synthesiser. It might be possible to support recorded clips as well as synthesised clips. Having on-demand generation of phrases is probably better (it can emphasise words better if it can understand the full context and raise the voice at the end of a question etc).
It would be good to have sign language (BSL initially) as a first class language option - so that would be video clips playing instead of audio output. It is also interesting to note that BSL has grammar rules that are probably more in line with what an AAC device user would naturally do, so "do you want a cup of tea?" would be signed as "Tea, You ?"