Open danyaljj opened 8 years ago
Also, would be great if someone trains SRL with jointTrainSparseNetwork and loss augmented inference. I guess it can improve the results. PR #445
Training for join is done and added to the documentation. If later anyone wanted to retrain all the models and do the renaming, please remove the property symbols also, for example:
val posTag = property(predicates, "posC")
should be
val posTag = property(predicates)
Do this for all properties.
Next time that you want to retrain SRL models, capitalize the name of the models in
SRLClassifiers
.