This program reads input in Constraint Grammar format, and splits special “multiword cohorts” into separate cohorts, leaving other cohorts and intervening blanks as they were.

For examples of input/output, see the files in test/, e.g. test/input.simple.cg.

A C++ compiler that goes all the way to 11.

Tested with gcc-5.2.0, gcc-5.3.1 and clang-703.0.29.

(Should work all the way down to gcc-4.9, but will fail with e.g. gcc-4.8.4 or clang-3.5.0.)

./autogen.sh
./configure  # optionally with argument --prefix=$HOME/my/prefix
make
make install # with sudo if you didn't specify a prefix

Takes no options, just stdin and stdout:

cg-mwesplit < infile > outfile

More typically, it’ll be in a pipeline after hfst-tokenise and some step that disambiguates multiwords using vislcg3:

echo words go here | hfst-tokenise --gtd tokeniser.pmhfst | vislcg3 -g mwe-dis.cg3 | cg-mwesplit

If you get

terminate called after throwing an instance of 'std::regex_error'
  what():  regex_error

then your C++ compiler is too old. See Prerequisites.