Apertium is a free/open-source rule-based machine translation platform. It includes both tools (compilers and runtime processors, mostly in C++) and language data (mostly in XML) for a number of languages, with a special focus on marginalised languages. Processing is typically done using finite-state transducers, with Hidden Markov Models or Constraint Grammar for disambiguation – individual language pairs differ in the tools they depend on.
There is a wealth of information on the Apertium Wiki at http://wiki.apertium.org
In particular, see http://wiki.apertium.org/wiki/Installation . If you are developing on language data, you'll want to use Tino's nightly builds: http://wiki.apertium.org/wiki/Install_Apertium_core_using_packaging