search

TALP-UPC / FreeLing

FreeLing project source code
Other
252 stars 96 forks source link

Documenting depenencies between modules #107

Open guludo opened 4 years ago

guludo commented 4 years ago

Hi there!

Thanks for this library! It is awesome.

I'm reading the documentation I'm having difficulties to know which morphological analyzer is required to use the senses module. I think it would be nice if the documentation pointed to the dependencies for a give module.

In my use case I just want to use the senses module, so I would like to enable only the required modules in the morphological analyzer.

Regards, Gustavo Sousa

arademaker commented 4 years ago

Hi Gustavo, I guess it will depend on your input format. How is your data? Is it a raw text? If so, you need to tokenize, split into sentences, and make basic morphosyntactic analysis: lemmatization, POS tagging. Keep in mind that multi-word expressions also need to be identified (Freeling can handle only adjacents MWE) and merged into a single token. Does it make sense?

guludo commented 4 years ago

How is your data? Is it a raw text?

Yes.

Does it make sense?

Yes! Thanks. :-)

Is there a page in the documentation listing dependencies between analyzers/modules?

lluisp commented 4 years ago

No, there is no such documentation

However, senses module requires only basic morphological analysis (i.e. "dictionary" module)

if you run freeling with default configuration, you'd get reasonable results for your goals. If you want senses for all PoS in a word, you need "--outlv morfo --sense all" If you want all senses on PoS tagged output, you need "--outlv tagged --sense all"

Any modules in the morphological analyzer can be deactivated at will, with --nonumb, --nodates, --noloc, etc (see user manual), depending on your needs.

As arademaker said, if you deactivate multiwords (--noloc) you may miss the sense for some multiwords existing in wordnet (e.g. "in_particular"), and get the sense only for each elements ("in", and "particular")