1 unstable release

0.5.0 May 15, 2022

#747 in Machine learning

MIT/Apache

53KB
1K SLoC

Oh No! More Lemmas

ohnomore consists of two tools to incorporate TüBa-D/Z style lemmas into language processing pipelines. The first tool, ohnomore-preproc takes TüBa-D/Z lemmas and transforms them into lemmas that are more fit for machine learning pipelines. For example:

  • Alternative lemmatizations are removed.
  • Separable prefix markers are removed.
  • Separable prefixes are removed when they are separated.
  • The special reflexive lemma #refl is replaced by the lowercased form.
  • Lemmas of truncations are replaced by their forms.

The second tool, ohnomore performs the opposite transformation (as much as is feasible).

Dependencies

~6.5MB
~112K SLoC