Relaxation Labelling PoS Tagger

The Relaxation Labelling PoS tagger only requires an appropriate pseudo- constraint grammar file, given by the RelaxTaggerFile option. See section 2.15 for format details.

To build a Relax tagger for a new language, you will need corpus (preferably tagged), and you will have to write some compatibility estimation scripts. You can also write from scratch a knowledge-based constraint grammar.

Nevertheless, the easiest way (if you have an annotated corpus) is using the estimation and smoothing script src/utilities/train-relax.perl provided in FreeLing package.

The produced constraint grammar files contain only simple bigram constraints, but the model may be improved by hand coding more complex context constraint, as can be seen in the Spanish data file in share/FreeLing/es/constr_grammar.dat



2008-01-24