multilingual ubiquity

[Pike]

Following our discussion on localizing ubiquity, I came up with
another set of scenarios, which are not so much about localizing
ubiquity to another language, but to deal with multilingual users.

I once overheard a pair of brothers, which were talking to each other
on French. At one point the conversation suddenly switched to English,
and I recognized that they obviously switched the topic to sports.
Guess those were Franco-Canadians, and SportsTV was on English, so for
sports, that was their native language.

For these guys, searching on the web might be "Recherche le web ..."
or it might be "Search for NBA results on espn" for sport results.

So for multilingual people, or people speaking minority languages (not
meaning franco-canadians here), the "native language" may depend
rather heavily on context.

Another scenario are migrants, for whom some key terminologie is not
in their native language. I have that experience with overhearing
conversations of young migrants in Berlin, which talk to each other in
their Mother tongue, but every now and then, they'd drop in a German
word, just because they never learned the name for that in their
native language, but only in German.

Thoughts?

Axel

["mitcho (Michael 芳貴 Erlewine)"]

Hi Axel,

Thanks for the interesting usage scenario... as you know, currently
you have to choose the language to be used in Ubiquity and only one is
loaded in at a time. Because the underlying parser is universal and
non-deterministic, however, it would be possible (to a certain extent)
to combine each language's properties at different points of the
parser and add a component to the score which requires (or strongly
prefers) that all the rules that were used in a parse be from the same
language. A simple case of this would be with verb names: we could
search for all the English verb names *and* the Japanese verb names
(for example) at the same time... if we recognize an English verb in
the input, then we could strongly prefer the parses which use English
rules for the rest of the parse. All of this, however, would explode
the already large number of intermediary possible parses to keep track
of and score.

I think this touches on the usability/learnability question of
Ubiquity in various langauges, though — there's a big question about
how we present this interface to users and shape their expectation of
what Ubiquity will be able to understand. If we show users an
introductory video in English, for example, would users feel locked
into having to use English? Alternatively, if we do a better job
presenting the user with Ubiquity in their locale, how should we
present the existence of commands which have yet to be translated to
their locale?

Thanks for the thought-provoking question — I'm looking forward to
working with you on these and other Ubiquity l10n questions!

mitcho

--
mitcho (Michael 芳貴 Erlewine)
mit...@mitcho.com
http://mitcho.com/
linguist, coder, teacher