(carmel includes EM and gibbs-sampled (pseudo-Bayesian) training)

(see carmel/LICENSE - free for research/non-commercial)

(see carmel/README and carmel/carmel-tutorial).

TODO

ln -sf . graehl;cd carmel; INSTALL_PREFIX=/usr/local make -j 4 install

(prerequisites: C++ compiler and Boost, which you probably already have on your linux system; for Mac, you can get them from Homebrew. Windows: you can use Microsoft's tools or cygwin or mingw.

If your system doesn't support static linking, make NOSTATIC=1

If you're trying to modify or troubleshoot the build, take a look at shared/graehl.mk as well as carmel/Makefile; you shouldn't need to manually run make depend.

• carmel: finite state transducer toolkit with EM and gibbs-sampled (pseudo-Bayesian) training

• forest-em: derivation forests EM and gibbs (dirichlet prior bayesian) training

• shared: utility C++/Make libraries used by carmel and forest-em

• gextract: some python bayesian syntax MT rule inference

• sblm: some simple pcfg (e.g. penn treebank parses, but preferably binarized)

• clm: some class-based LM feature? I forget.

• cipher: some word-class discovery and unsupervised decoding of simple probabilistic substitution cipher (uses carmel, but look to the tutorial in carmel/ first)

• util: misc shell/perl scripts