The issue is in fact with the English - Arabic pair. The amount of
effort needed by Google for "statistical" translation goes as n(n-1)/2
where n is the number of languages. I am beginning to feel that the
strategy of Google is unsound. Kurzweil talked about Google having AI.

http://www.crn.com/it-channel/173603287;jsessionid=4AVRUOF5SDBFNQE1GHPSKHWATMY32JVN

There is clearly very little that could be described as AI in Google
Translate.

Latin is a language that makes distinction in terms of ending between
subject and object. If I were to say "Colonus Palestinium
interfecit" (I would in fact use the word "Colonus = settler" instead
of Israeli) the endings have a significance. As GT does not look at
grammar it isn't clear how it would cope with (in particular) Latin to
English. If I were emphasising that it was a Palestinian who was
killed I would say "Palestinium colonus interfecit" the "um" and the
"us" tell us who is killing who.

http://groups.google.com/group/google-translate-general/browse_thread/thread/7fe883bd287c1cf4?hl=en

I have said that statistical translation -> n(n-1)/2. Yet the effort
involved in humans learning languages is in fact < n. Many people
start learning Arabic at university having never studied it before.
Schools of Arabic though tend to insist on people having a good
standard in a second language. People tend to learn Arabic quicker
than they learn Spanish. As I said the reverse is the case for GT that
treats each pair as a separate case. This paragraph proves that
Kurzweil was talking though his hat and there is nothing recognisably
AI about GT.

As I said in Latin grammatical distinctions are the meat of it and it
is hard for me to imagine satisfactory Latin from a system that by its
own admission does not invoke grammar.

GT has been hacked into. Hilary Clinton was telling the Chinese
(rightly) that hacking was against International Law. The same surely
applies to Israel. However this is not really the point. The fact of
the matter is that statistical system are easier to hack than
grammatical ones and tend to fall over as the number of languages
increase. They are in no respects AI.


  - Ian Parker

On Jan 26, 11:24 am, Pertinax wrote:
> I think this might be an Arabic issue.
>
> For instance translating "I would like to buy a mat" into French
> produces
> "Je voudrais acheter un tapis", which is exactly what I would say.
>
> Similarly, in the English to German translation, "Ich möchte eine
> Matte kaufen".
>
> As you can see, neither has a preposition in front of the infinitive.
>
> Quod erat demonstrandum.
>
> On Jan 25, 10:24 pm, Ian Parker wrote:
>
>
>
> > Amare means to love. It is an INFINITIVE. I did Latin ant school and I
> > simply cannot visualise Google producing a good Latin translation.
>
> > I translated "I would like to buy a mat" into Arabic. I got
> > وأود أن شراء سجادة
> >  أن is an English issue not an Arabic one. If English was sent
> > through
> > CLAWS this would have become apparent. "to buy" is an infinitive.
> > This
> > issue affects all languages not just Arabic.
>
> > Volo equum mercari NOT Volo ad equum mercari as Google would insist.
> > أن is of course the Arabic for ad. How can you possibly hope to
> > translate into Latin when your knowledge of basic English grammar is
> > go ropy. As I said أن is an English and all language issue NOT an
> > Arabic one.
>
> >   - Ian Parker
>
> > On 25 Jan, 19:03, Bud wrote:
>
> > > We need to include Latin among the language groups.

Reply via email to