To use this language-pair package with apertium YOU DO NOT NEED TO
RETRAIN THE TAGGER. Probabilities and auxiliary data are provided for
both the pt-es and the es-ca translation directions which should be
acceptable for most applications.

If for some reason you need to retrain the tagger (for example, you
have made extensive changes to the dictionaires), you will need to
assemble a large (millions of words) plain-text corpus for each
language (for example, using a robot to harvest text from online
newspapers), and run it through the morphological analyser (lt-proc)
to generate a file with name composed of the language pair codes and
the extension .dic and make sure you have a copy (or link to it) with
extension .crp (for instance pt-es.dic and pt-es.crp). In subsequent
versions of apertium, a tool processing the dictionaries will
circumvent the need for this large corpus.

If this sounded too complicated, that's because it is. You don't want
to retrain until a later version of apertium is released. Believe us.
