This model is fine-tuned for historical bibliographical named-entity recognition
AjMC-NE-Corpus, so as release in
HIPE-2, for 40 epochs, beating all our other models.
It is intended to detect persons (authors, mythological...), works (primary and secondary literature, fragments), objects (manuscript, inscriptions), locations, dates, and scopes (i.e. reference scopes, such as Hom. Il.
XI, 106-122).
It can be used in a normal token classification pipeline.
