From: Brian Wolven <brian.wolven_at_verizon.net>
Date: Thu, 08 Apr 2004 13:03:49 -0400

[IDC]Dragon wrote:

> It definitely won't fit into something like the current voice file. But
> maybe some kind of caching could be done for it.
>
> I'm afraid the number could be higher than your amount of mp3 files. And we
> need a good definition what a "word" is (e.g. delimited by non-alpha chars).
> English gives no lower bound on the char count, since "a" and "I" are words.
> But not every group of chars makes sense as a word, should rather be spelled.
> A collection of digits could form a number then.

Some filtering of the collected word list with regular expressions could
handle most of the exceptions, I think. That introduces more complexity
in the code that decides which words need to be 'read', however. There
is probably a workable middle ground somewhere in there. Whether it is
feasible to implement this on our Archos hardware is another question.

> The interesting part about it is that also the files could be reasonably
> voiced. The bad part is a lack of consistency. When you delete files/dirs, the
> words remain. The maximum directory size will give trouble for the "words"
> dir, if all are separate.

Perhaps breaking the word dir structure down alphabetically? Maybe I'll
run a test on my own music collection, just to get some
order-of-magnitude idea.

> Probably there's a lot more complications when thinking further about it,
> something I'm too lazy right now...

Lazy? Heh. Right...

_______________________________________________
http://cool.haxx.se/mailman/listinfo/rockbox
Received on 2004-04-08