[FoRK] Slow grep

Gavin Thomas Nicol gtn at rbii.com
Fri Mar 19 08:01:11 PST 2004


On Friday 19 March 2004 09:48 am, Eugen Leitl wrote:
> On Fri, Mar 19, 2004 at 09:14:17AM -0500, Gavin Thomas Nicol wrote:
> > Except that human language is richer phonetically than that implies, and
> > also
>
> No, learning some 20 glyphs and understand that they map to some noises
> (as long as you don't have to pronounce them all is peachy), not concepts
> is well within a computer user faculty envelope.

I would debate that point. How would you distinguish tchi, shi, tschi, for 
example? Far too easy to create ambiguitites.

> > has contextual dependencies. For example, if I say "Ikura wa ikura" you
> > have
>
> For example, if I write "mama" in English, it's pretty clear I'm not
> talking about horses. Neither clipboard nor keyboard care too much about
> the phonetics.

Right, but phonetics alone are insufficient, that was my point. This is 
especially so if you have a (native language)->phoneme->(native language) 
mappings with largish numbers of homonyms, and it becomes even worse when you 
have cultures that *enjoy* creating ambiguous phrases, which many cultures 
with richer languages than English do.

FWIW. This is more than idle speculation on my part... I have watched Japanese 
people trying to read texts that were all Hiragana (40+ phonemes), and they 
have trouble (and for me, it was *really* hard). Imagine a computer doing any 
better?




More information about the FoRK mailing list