cygwin?

[brian]

Hi,

First I wanted to say thank you for making this available. This is a
huge benefit to getting me up and running with moses.

I wanted to ask if the scripts had been tested under cygwin (I read
there had been issues with different versions of Ubantu)?

Thank you!

Brian Wolf

gOgO development, ltd
sedona, az

[Hieu Hoang]

not sure what this moses-for-mere-mortals is but good luck with it

i personally don't test the training scripts on cygwin, but i know a few
people have it running and send patches to keep it running occasionally.

more info here
http://www.statmt.org/moses/?n=Moses.FAQ#ntoc8

[Moses.for.Mere.Mortals]

Hi, Brian,

Sorry for my late answer.

No, the scripts have not been tested under cygwin. They have just been
tested with Ubuntu 9.04. I had problems compiling several components
(e.g., IRSTLM) under Ubuntu 9.10 (I haven't yet given up, but I'm
waiting for some more updates or info about 9.10). For the time being,
I'm more concerned with increasing the functionality of what already
works in Ubuntu 9.04 and there should be some fresh news pretty soon.
A person that seems me particularly interested in cygwin, and with
whom I have already exchanged some e-mails, is Ivan Uemlianin (see,
for instance, http://article.gmane.org/gmane.comp.nlp.moses.user/2342/match=cygwin
).

My objective is to build a prototype of a translation system that can
be used in practice by professional translators (that is, not
necessarily informatics experts, but still enthousiastic persons who
are acquainted with several MT systems and in some cases participated
in the linguistic side of their development). In fact, I was contacted
by a small group of such friends who were having problems compiling
Moses and the packages upon which it depends (and yes, they are using
Ubuntu 9.04 and I also had problerms to start with). I found my work
could be useful to even more persons. It has already been used for
practical translation work with a trained corpus of several hundreds
of millions of words (created by them; it surely determined the
results) and the first results were encouraging (users compared the
system to Google and Systran). Still, I know there is much more that
can be done, including many things that we still have to learn (always
a nice perspective). Suffice it to say that, to start with, I didn't
believe *at all* that anything useful could come out of SMT, though
they managed to keep me working ;-)

Best regards and a nice 2010;

João (Rosas)

[brian]

Hi Joao,

Thanks for the kindly response. I will keep this on the "back burner",
will definately be back in touch after I take a closer look (might be a
while). On one level, I simply would like to learn more about the
subject. It does sound very interesting. I am a programmer, although
without much NLP background, I am learning along the way. I work for a
small company that does IR work, but am also a computer and technology
"hobbyist".As a "hobbyist", I am also curious about a variety of things
as code and memory optimization, parallelization, potential applications
to IR, ie new multi-lingual search algorithms, etc, as well as mobile
apps, speech recognition, data mining, web marketing and seo type
things, and new types of text generation (ie source and target are the
same language).

Happy New Year to you too!

Brian

> Jo�o (Rosas)