The heart of VIDasher

In VIDasher, you generate one word at a time. The switch actions required to make a particular word are always the same, so it's possible for you to gradually memorise the VIDasher codebook.

You use short presses to navigate down the tree of words. The two short button presses indicate `earlier' and `later' respectively. (In the linux prototype, `earlier' and `later' are produced by the keys "0" and "-".) The guess is presented to you either visually or by audio (text-to-speech) or both.

A long press of the `later' button indicates `that's right!' (In the linux prototype, `that's right' is produced by "=".) Once you have said `that's right', you immediately start navigating the next word.

If you are part-way through navigating into a word, a long press of the `earlier' button undoes the latest `earlier' or `later' press. (In the linux prototype, `undo' is produced by typing "9".) You can reverse all the way back through the tree to the beginning, but you cannot undo a `that's right'.

The above system allows you to write a string of words and to correct errors in the current word.

If you want to be reminded what you've written and where you are, press any other key to get a status message. [If there is demand for this feature, I'll figure out how to obtain the status message using only the main four actions; we could use the sequence 909 or 9-9, for example, to enable the status message for two steps.]

Personalization

VIDasher reads in a dictionary of words with frequencies (represented by numbers bigger than 1). You can define this dictionary in whatever way you want. We recommend that the dictionary should be alphabetically sorted (so you can navigate round it) and that your important and favourite words should be given big frequency. To help you make an appropriate dictionary, a program, personalize.p, is supplied which finds the frequency of words in any text file. To make your favourite words more probable than dictionary words, whose counts (in my dictionary) go up to 8000, I recommend scaling up your personal word counts by a factor of 1000. [This is the default factor in personalize.p.]

In the prototype of VIDasher, the dictionary is made by merging and sorting a selection from the following files:

mydict

- every word in the linux spellchecker (84,000 words) has been included (stripped of punctuation characters such as apostrophes) with a frequency that's biggest for short words and smallest for long words. For example, four-letter words have a frequency of 1024, and words like editorializing get a frequency of 1. The spellchecker includes rude words. You might want to filter these out.

polite

- the same as mydict, but with some rude words removed.

emmadict

- 7,000 words from Jane Austen

alphabet

- the 26 characters of the alphabet (1000000 each)

radiospell

- the words alpha, bravo, charlie, ... (1000 each)

generaldict

- counts produced from the dasher training text. (every word gets 1000 for each occurence in that file) (8600 words)

words

- some more words that I wanted to throw in or tweak (123 words)

yesno

- the words yes and no -- but see below for another way of handling these special words.

phrases

-- 122 words used in my favourite phrases.

It's fine to include words twice, as long as the whole dictionary has been sorted alphabetically.

Punctuation and Numerals

We can include punctuation characters and numerals and number-word combinations such as 23rd or 9.30am in the dictionary.

'numbers'

supplies numerals, numbers, and number-word combinations (with frequencies up to 250)

`punctuation'

can be prepended to a dictionary to supply a few punctuation characters if these are desired.