[Kde-accessibility] KTTSD - language detection

Mikolaj Machowski mikmach at wp.pl
Sun Jan 30 21:29:04 CET 2005


Hello,

It's me again :)

Talker Chooser works but language detection made with it is primitive.
I've set it to checking for Polish diacritics but they are not so common
to make this detection comfortable successful.

I have an idea how to make it automatic and working most of the time.

Idea is borrowed from VimSpell by Mathieu Clabaut (working on suggestion
by Hugo Haas) - this is always working for me.

This is how it work:

Pass text to spellchecker, check against languages which voices are
installed in KTTSD and choice voice of language with lowest number of 
errors.

For short texts and in slang environments like Kopete it probably fails.
But should work with mail, web pages or longer texts in other
applications.

It's kinda weird to use spellchecker for setting voice in t-t-s app but
should work :)

Let's take scenarios for bi-lang person:
1.
Installed two dictionaries for spell-check and voices for two languages.
Easy. Check spelling for both and choose this with smaller number of
errors.

2.
Installed one dictionary.
Decline automatic detection without all dictionaries installed.

I thought can come with nice guessing scheme but it is not easy.

If person is using more than one language he/she will have
installed spellcheckers for all languages - IMO this is safe bet. Better
than wild guessing depending on human characters.
  
m.

-- 
LaTeX + Vim = http://vim-latex.sourceforge.net/
Vim Universal Templates 
	http://www.vim.org/scripts/script.php?script_id=1078
CLEWN - http://clewn.sf.net




More information about the kde-accessibility mailing list