Jackson Petty and Bob Frank present at the Workshop on Computational Models of Reference, Anaphora and Coreference
Linguistic major Jackson Petty ‘22 and Bob Frank presented a talk on models of learning for anaphoric dependencies at the 3rd Workshop on Computational Models of Reference, Anaphora and Coreference (CRAC). Jackson and Bob’s work explores the conditions under which recurrent neural networks can learn a lexically abstract meaning for reflexives, i.e., one which establishes co-reference with any structurally appropriate name, even one that hasn’t been previously encountered. Such abstraction had previously been thought to be beyond the capacity of neural networks. The CRAC workshop was part of the 2020 COLING conference, which was to be held in Barcelona, but took place virtually because of COVID-19.