Morphological analysis for Spanish

[olzama]

The current dependency on Freeling is expensive to maintain and, while providing some robustness for unknown words, is also restrictive when a wrong tag gets output.

Berthold suggests implementing morphology directly in the grammar. Unknown would still need to be handled by some statistically trained tagger.

If we keep Freeling, perhaps some kind of two-stage procedure is needed such that, if a sentence is not parsed, it should be tried again, this time not only with Freeling best tag(s). For verbs it is perhaps not possible but many words can be found in the lexicon directly, no morphological analysis is needed?