Hi Ken, You might consider starting with one of the packaged distributions of Moses to get you through the hump. There are several to choose from.
http://www.statmt.org/moses/?n=Moses.Packages Our distro is DoMY CE v2.5. Version 1.0 was released in November 2010. The current version has Moses beta 0.91 and we will upgrade to RELEASE 1.0 in the next few weeks. It also includes all source code so you can compile changes to Moses' kernel, and scripts to integrate those changes to your runtime if you like. Tom On 03/09/2013 08:55 PM, Philipp Koehn wrote: > Hi, > > you seem to have the right background and have made the right > first steps! > > For the practical use of Moses you will not fully understand all the > intracies of the underlying mathematical models such the details > of EM training of IBM Models in GIZA++. And the dominant model > (the phrase model) is not very linguistically sophisticated. > > If you have to deal with very large corpora, I'd suggest to take > a closer look at the pruning and compact representation methods > described in the advanced feature section: > http://www.statmt.org/moses/?n=Moses.AdvancedFeatures > > -phi > > On Sat, Mar 9, 2013 at 9:52 AM, Ken Fasano <[email protected]> wrote: >> My task, at work, is to learn Moses and SMT well enough so that in a year or >> so we can leverage a huge (billion word) multi-language corpora (mostly W. >> European, but possibly also Chinese, Japanese, Korean) to translate across >> industries (electronics stores, transportation, government, etc.). I'm the >> only one in the company working on this. I am starting out with 30 yrs. >> experience as a software developer, so Linux, C++, Perl, etc. are old >> friends. My linguistic background is almost nil, except for the ability to >> read German, French, Spanish, Italian and a little Greek, and understand >> their grammars, as well as concepts such as CFGs. My major in college (years >> ago!) was music composition; my sole linguistics course was IPA phonetics >> (the composer Stockhausen told me to take it!), and the nearest I've been to >> the math involved in SMT is Markov chains (thanks to the composer Xenakis). >> I've been playing with Moses, going through the tutorials, reading as much >> on the web as I can, understanding some of the concepts but freaked out by >> the math. I've ordered Dr. Koehn's book; it hasn't arrived yet. >> >> I'd deeply appreciate any tips on how to get from HERE to THERE! >> >> _______________________________________________ >> Moses-support mailing list >> [email protected] >> http://mailman.mit.edu/mailman/listinfo/moses-support >> > _______________________________________________ > Moses-support mailing list > [email protected] > http://mailman.mit.edu/mailman/listinfo/moses-support _______________________________________________ Moses-support mailing list [email protected] http://mailman.mit.edu/mailman/listinfo/moses-support
