Apertium lttoolbox, master is svn trunk, stuff happens in branches. The branch lt-trim_fstp_new_transducer seems to work for both regexes and regular entries; still TODO are compounds and <j> and <g> multiwords (<g> could get ugly, the others should be simple). The branch lt-trim-2transducers doesn't use FSTProcessor for analysis, would have to reimplement some parts of it, but might end up being cleaner in the long run.
README FILE OF LTTOOLBOX
1. Programs
lt-comp -> compiler, execute without parameters to show the
instructions of use.
lt-proc -> processor, works with options -a (lexical analyser,
default option), -g (lexical generator) and -p
(lexical post-generator). Using -h will show all
flags.
lt-expand -> generates all the pairs of transductions of a given
dictionary. Execute without parameters to show the
instructions of use.
2. Install
a. Requirements:
- i686, ppc, SPARC, etc.
- g++ >= 2.95
- gnu make
- libxml2 (last version)
b. Building & installing
- ./configure
- make
- make install (o make-install script)