Google Translate

[Warren Smith]

Some machine translation problems are more egregious than others....

Put 48億円 into Google Translate (which I sometimes use as a dictionary) and
you get "48 billion" rather than 4.8 billion. Ouch!

Warren

[Chris Nichols]

Howdy Warren,

In the immortal words of Groucho Marx:

"Well then, don't do that." (Relating advice from his doctor)

Cheers,

Chris Nichols

[David J. Littleboy]

From: "Warren Smith" <warren...@comcast.net>

>>>>>>>>>>>>>>>
Some machine translation problems are more egregious than others....

Put 48億円 into Google Translate (which I sometimes use as a dictionary) and
you get "48 billion" rather than 4.8 billion. Ouch!
<<<<<<<<<<<<<<<

Hey! I make errors like that all the time. That's why I have our CEO
proofread all my work before I deliver it. Useful things, CEOs. Highly
recommended.

If you read the popular press, though, you'd learn that computers are making
great strides.

But in real life MT is getting worse, not better.

Why? Because the MT types aren't even trying to get it right, and don't care
if they get it wrong. Trying to get it right is too hard*, so they look for
easy (computational) ways out. Which are usually intellectually vacuous,
like "corpus-based methods". For starters, any corpus will be full of
errors, of course.

*: Back in the early days, the MT types tried to have their machines
actually parse the input and figure out what was going on. But no more.

David J. Littleboy
Tokyo, Japan

[S Zaveloff]

Interestingly enough though, if you enter 4億円, you get 400 million yen.
Other single digits are also correct. However, multiple digits and no
digits (i.e., 億円) come back with billion yen.

--
------------------------------------------------------------------------
Steven H. Zaveloff gua...@gmail.com
P.O. Box 200203 Tel: (512)219-7142
Austin, Texas 78720-0203 Fax: (512)233-2770
http://members.capmac.org/~stevenzaveloff/

Thus shall you think of this fleeting world:
A star at dawn, a bubble in a stream;
A flash of lightning in a summer cloud;
A flickering lamp, a phantom, and a dream.
-Diamond Sutra

[Tom Gally]

四十八億円 yields "84 million yen."

Google Translate seems to be dyslexic about 漢数字 in general:

パーティーに二十人が来ました。
12 people came to the party.

Tom Gally

[Jon Johanning]

I really cannot understand why human beings who are supposed to be
intelligent expect "artificial intelligence" to come out of computers,
which are manifestly no more intelligent than an electric pencil
sharpener--just more complicated. The idea seems to be that if you make
them really, really, super-complicated, they will get smart by some
mysterious transmogrification.

Also, have they forgotten "GIGO"? Try to build up a "computer
translating system" by hoovering up huge amounts of allegedly properly
translated texts and manipulating them statistically, forgetting that X%
of these texts are going to be full of mistakes, and what do you get?

"Computer translating system"? Get back to me on that when you have
invented a machine that understands language the way we humans
understand it. (Hey, we invented language in the first place.)

Well, back to work as a humble, human translator.

Jon Johanning // jjoha...@igc.org

[Kamana Akina]

The way google does things in terms of translation is a bit different, they used what's commonly referred to as a statistical analysis, meaning instead of people at google actually putting in translations or grammar structures or any specifics of a language whatsoever, they use search results from the internet and pull out the most commonly used instance of a certain word in another language. This explains why certain pages, like fillable forms, when translated using google translate, tend to yield translations that make more sense, since they are using more commonly-used words. Even so, most of the time it still puts out horrible english equivalents that would make even college students studying Japanese cringe. But still, compared to a few years ago the translations seem to have improved, and seeing as this method is largely content-driven, I can see it improving more and more further down the line. but of course, could it replace a human-translator? most definitely not.

--
You received this message because you are subscribed to the Honyaku Mailing list.
To unsubscribe from this group, send email to honyaku+u...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/honyaku?hl=en?hl=en

[Jean-Christophe Helary]

On 16 juil. 10, at 04:19, Kamana Akina wrote:

> could it replace a human-translator? most definitely not.

I agree, but it will profoundly modify the job market structure (more bilingual proofreaders/rewriters) and most certainly will expend the market itself, especially in combination with the development of the e-reading/e-writing market.

Jean-Christophe Helary
----------------------------------------
fun: http://mac4translators.blogspot.com
work: http://www.doublet.jp (ja/en > fr)
tweets: http://twitter.com/brandelune

[David J. Littleboy]

From: "Jean-Christophe Helary" <jean.christ...@gmail.com>

> On 16 juil. 10, at 04:19, Kamana Akina wrote:
>
>> could it replace a human-translator? most definitely not.
>
> I agree, but it will profoundly modify the job market structure (more
> bilingual proofreaders/rewriters) and most certainly will expend the
> market itself, especially in combination with the development of the
> e-reading/e-writing market.

I think the market getting expended is something we all should fear...

David J. Littleboy
dav...@cheapshots.com
Tokyo, Japan

[Jean-Christophe Helary]

Expand.

[Brian Chandler]

Jon Johanning wrote:
> I really cannot understand why human beings who are supposed to be
> intelligent expect "artificial intelligence" to come out of computers,
> which are manifestly no more intelligent than an electric pencil
> sharpener--just more complicated.

Perhaps because presumably "natural intelligence" is supposed to
emerge from your brain, the neurons making up which being manifestly

no more intelligent than an electric pencil

sharpener--just present in vastly larger numbers. (See Searle's
notoriously false "Chinese room")

So your argument, and the slur on people working in AI, is invalid.

> Also, have they forgotten "GIGO"? Try to build up a "computer
> translating system" by hoovering up huge amounts of allegedly properly
> translated texts and manipulating them statistically, forgetting that X%
> of these texts are going to be full of mistakes, and what do you get?

You get junk, obviously. Which is exactly what you get when you treat
humans as sausage-machine translating devices. And the machine-
generated junk is a lot cheaper, which if you are a Business Mind is
all that matters.

Brian Chandler

[Simon Varnam]

Kamana Akina wrote:
>
> The way google does things in terms of translation is a bit different, they
> used what's commonly referred to as a statistical analysis, meaning instead
> of people at google actually putting in translations or grammar structures
> or any specifics of a language whatsoever, they use search results from the
> internet and pull out the most commonly used instance of a certain word in

> another language......

They also have a little box a the bottom of the page for you to suggest a better translation.
Free labour?

I wonder how much extra weight they give to these human translation over the mass of stuff filtered from the web.

Simon Varnam

[Irina Knizhnik]

Simon Varnam wrote:

They also have a little box a the bottom of the page for you to suggest a
better translation.
Free labour?

Free (translator) labor is the name of the game.
Irina Knizhnik