Re: Italian News: Fed to play exhibition pre Wimbl.

On Jun 17, 10:41 am, "arnab.z@gmail" <arnab.zah...@xxxxxxxxx> wrote:
On Jun 17, 5:45 am, Javier Gonzalez <jagon...@xxxxxxxxxxxxx> wrote:





Joe Ramirez <josephmrami...@xxxxxxxxxxx> wrote:
So how does the software choose? As a tier I analyst would say,
"context" - and now you are asking for a computer that can reliably
understand human-language context. No such thing exists to date :(

Doesn't have to be perfect understanding -- just better than poor,
which is what it is now.

Trust me, "poor" is quite a challenge ;)

How about because the word "Wimbledon" is in there?

the championship? or the movie? or the suburb? or the football team? the
theatre? the speedway? the wine company? the running club? ;)

As an application gets more general-purpose, it loses effectivity in
particular cases. You could have an implementation that is
tennis-specific (e.g. knows that "shot" is a tennis stroke and not a
bullet fired from a gun or a small glass of distilled booze) but that
would not serve the google translator's purpose, which is to provide
completely domain-agnostic automated translation - and it doesn't get
any more general than that. And even with that specialization, you can
run into cases that won't be resolved properly by the computer. For
example, you could feed it an article about the movie Wimbledon, which
does relate to tennis, but you then could have shot as in tennis stroke
and shot as in framing and filming, and the need for a machine that
understands context strikes again.

As I said before, automated translation is a *very* complex matter.

With current technology, near perfect translation requires a large
amount of world knowledge to aid in disambiguation of context. Even
humans cannot parse everything correctly (97% precision is the human
gold standard). That means we need tons of distributed database
servers holding terabytes of real world textual and spoken data.
Moreover, the cognitive faculty of language inside the human mind is
hard to replicate with current technology. Chomsky alone, for the past
40 years, directed linguistics towards creating a symbolic, formal,
category-based linguistic theory and thwarted the growth of
probability-based, nuanced, graded account of human linguistic
competence. After the AI winter, things became even slower. With
hardware getting faster, bigger and cheaper every year, we are now at
a stage to tackle the language problem with a lot of force. And Google
is the biggest giant in this field, "collecting all the information in
the world", as they claim. If any company is in a position to make
machine translation possible through sheer redundant statistical
weightage, it's probably Google.

Yes, I think that if Google gave its translation software the same
attention it gave its search engine, the translation tool would be
excellent, as opposed to mediocre-to-lousy.

Joe Ramirez


.

Relevant Pages

  • Re: Italian News: Fed to play exhibition pre Wimbl.
    ... understand human-language context. ... automated translation is a *very* complex matter. ... the cognitive faculty of language inside the human mind is ... 40 years, directed linguistics towards creating a symbolic, formal, ...
    (rec.sport.tennis)
  • isa 7:14
    ... Quote:Originally Posted by username What I noticed was that you are ... Context is very important. ... That's why one should disregard the virgin ... the intent of the various translation committees) can be determined by ...
    (alt.religion.islam)
  • Re: Yet more crypt cards
    ... > one message in the tree. ... that sounds like a big drawback to your reader. ... > It's how Google works. ... Or that the onus for digging out context should ...
    (rec.games.trading-cards.jyhad)
  • Re: Giving an application a window icon in a sensible way
    ... nothing was altered or taken out of context. ... I provided a link to the Google page showing the ... googling at all. ... And the even bigger problem is that you automatically ...
    (comp.lang.java.programmer)
  • Re: "Kouya Ruten" - Song Translation - Full Version
    ... After rereading the translation in light of your comments, ... However far, along this never-ending path ... Like a tale told in a dream, the ship of time sets out. ... Doesn't the context just meet when you use 'however' here? ...
    (sci.lang.japan)
Loading