search

direct-phonology / phoNy

phonology in spaCy!
MIT License
0 stars 0 forks source link

create model architecture for tagging separate syllable components #4

Closed thatbudakguy closed 2 years ago

thatbudakguy commented 2 years ago

see https://thinc.ai/docs/api-layers#multisoftmax

thatbudakguy commented 2 years ago

it looks like all we'd need to do is borrow https://github.com/explosion/spaCy/blob/master/spacy/ml/models/tagger.py and change the Softmax Layer to MultiSoftmax. then all of the configuration happens on the component, not on the model, since the model (called something like MultiTagger, maybe) only cares at runtime what the output vector sizes are.

thatbudakguy commented 2 years ago

probably should depend on #10