Mathieu Tozer's Dev Blog

To solve the problem of the computer getting to know what words you
know and thus don't know.
You have to distinguish between two or three ways:
1. Users completely learning from scratch.
2. Users who already know some words.
or
3. Users who know a little bit.

In case 1, you can afford to assume that the user, and the computer know nothing. Learning Finnish, for example. Therefore the original idea of building a list of words
that the user knows and doesn't know will work. This will be the case
in many cases, however as in

Case 2, the user is a native speaker or knows a bit already, and
therefore the computer wouldn't be able to be brought to that level
without a whole lot of kerfuffle, which would ultimately drive the
user away. So the solution I thought of while in the bath is to have
the user indicate the unknown words with a double click copy kind of
thing, so that the computer can know at least something about what
the user knows and doesn't know. The entire text can still be parsed,
and only the words that were clicked will be added to unknown, the
others to known. This should happens in both case 1 and case 2,
although in case 2 it is more integral to the solution (in a subtle
way).

So 一から users are assumed to know nothing, all words are new.
All other users are assumed to know everything, and required to
indicate the words they don't know. The words they don't indicate,
are assumed to be known.

Users should be asked what case they are during language setup, so
that Words knows how to handle text input, and how to manage the
dictionary.

After time, a text will automatically show words you know / don't
know, and then the list will refine as you click on individual words.

--

The above functioning must be thought out well, as it is very
important functioning for Words. Slowly I am ironing out the
difficulties. I think with this latest bath-think, I have tackled the
problem of getting the computer to know you.

--

It's important to distinguish between daydreams that are realizable
and those that are fantasy.

Words will function something like a 'history list' for a dictionary but will be more dynamic.

There will also be the possibility that languages without an existing word parser can be learnt with Words because of the single word feature. Then the software would just be used for unifying and managing and learning vocab lists, but that's still pretty good.

Parsers for Japanese and Chinese could be developed with a 'try and then see what you get' kind of algorithm. Much like the way I check these two, then that one, then the preceding one. And then look down the list of things to see what might fit.

So when a language is selected from the list that a parser or dictionary does not exist for, the program should behave appropriately, informing the user what functionality will be available.