I have one suggestion to you. You should demand that Arabic be
included in MOLTO. You should contact Arne Ranta at Stockholm. I have
had some long stating criticism of GT in general terms not just
specific languages. Arabic has no verb "to be". In English
translations this should be inserted. In Russian. I don't feel when
looking at GT we should be concentrating on just one language.
Russian has an attributive Девушка красивая = The girl is beautiful.
This is in fact wrong. In the attributive the final я is missing hence
Девушка красива. красива also means red. The fact that genders are all
over the place (French, German, Russian, Arabic) and no grammatical
cognisance is given is a major criticism. Many Arab countries are in
MINA so although Arabic is not an EU language it is very close to
being so. Closer in fact than Russian which IS included in MOLTO.
To translate Arabic satisfactorally cognisance must be taken of
grammar. I an a retired scientist and I have been writing a
translation program (Arabic - English) in my spare time. It decomposes
a word into prefixes and suffixes and lists all possible meanings. It
is in fact based on Tim Buckwalter's dictionary and does grammatical
checking based on morphology type. If a settler kills a Palestinian
for example it will simply treat each word as it comes. English will
be stilted but will have the correct meaning.
Another Area where GT fails spectacularly is in countries. Since the
end of the Vietnam war the US has not been involved in SE Asia. Yet US
troops committed atrocities in Burma. This was I think Vietnamese.
MOLTO to my way of thinking provides the only sound answer. It is
designed to be truly multilingual. It has to put every language into a
scheme by which every language can be translated into any other. What
you need for MOLTO is the translation of Arabic into some sort of
common base.
I can (at the moment) translate Arabic semi automatically. That is to
say I can generate a file with a C++ program which is read by a Java
program where manual selections can be made. In a fully automatic
translator these options will have to be evaluated using Latent
Semantic Analysis.
It should be pointed out that the "common basis" of MOLTO includes Web
3.0 constructs. Including OWL.
Consider :-
اما بالنسبه لأحجام النجوم فيتم التعرف عليها من خلال تطبيق قانون ستيفان
بولتزمان في الاشعاع (AmA bAlnsbh lIHjAm Alnjwm fYtm AltErf ElYhA mn
xlAl tTbYq qAnwn stYfAn bwltzmAn fY AlA$EAE) I have put the Buckwalter
transliteration in.
مضروبا في اربعة اضعاف درجة حرارة سطحه اى (mDrwbA fY ArbEp ADEAf drjp
HrArp sTHh Ay
Anyone who understands any Physics knows that the Stefan Boltzmann law
is the 4th power of absolute temperature. GT has 4 times the
temperature. This is dimensional nonsense. You see قانون ستيفان
بولتزمان is an OWL statement. MOLTO will vector me into statements
about the law. You see translation is about hyperlinks as well as
words. OWL will, of course, give me the correct law. This is starting
to take us deeply into AI. Can you translate well without
understanding? I am in fact convinced you can't. You can do a damn
site better than GT, but perfect no.
درجة is ignored by GT completely, so it would be a mistake to say
that you needed understanding for that sentence. Understanding (a
hyperlink) would provide a check though. درجة in this context can be
freely translated as "power of".
This common base then is also a springboard for AI.
In Vietnamese (and Arabic as well) countries are frequently
mistranslated. The remedy for this is simple. In our word matching
schema we can reduce "United States" to "{country}". We then translate
in two states. We first translate Arabic into generic English and then
resolve the generics. Tis is trivial. Will Google do it? No way.
I think you really need to start from scratch.
There is one thing about Google though. It does have a large database.
I can augment my Buckwalter vocabulary using GT. However I cannot get
a morphology type. Could you suggest any Arabic dictionaries in
Unicode?
- Ian Parker
On Jun 29, 7:24 am, Dr. Tahir Aletewi wrote:
> To Google,
> Regarding the translation from Arabic to English for the following
> statements that appeared in a journal early this month...
> ÇáÕåíæäí áÇ íÍÈ ÇáÓáÇã
> ÇáÚÑÈí áÇ íÍÈ ÇáÓáÇã
> Which means:
> - The Zionist doesn't like peace
> - Arab doesn't like peace
>
> Amazingly your translator translated the sentences as follows
> - The Zionist peace-loving
> - Arab does not like the peace
>
> when I copied the sentences and did a reverse translation (English to
> Arabic) for the results above I got the following...
> ÇáÕåíæäíÉ ÇáãÍÈÉ ááÓáÇã
> ÇáÚÑÈí áÇ íÍÈ ÇáÓáÇã
> Astonishing result, The first sentence has been totally reversed in
> its meaning... I did submit a correct translation but still have the
> same result... I have noticed that another person also submitted a
> correction as shown in your forum but it still give the same
> mistake...
>
> I guess it is an honest mistake or maybe someone has suggested such
> translation due to the problems in the area... I hope that you will
> correct the translation ASAP since there is an email that is being
> circulated about this....
>
> I'd be happy also, to review any Arabic to English or English to
> Arabic suggestions in the future
>
> Regards,
> Dr. Tahir Aletewi
> IT Faculty - AHU University
> [email address]
> [email address]
--
You received this message because you are subscribed to the Google Groups
"General" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/google-translate-general?hl=en.
