Bob Frank receives NSF grant to study inductive bias in neural networks

Bob Frank has been awarded a grant by the NSF on the topic of “Inductive Biases for the Acquisition of Syntactic Transformations in Neural Networks.” This work, in collaboration with Tal Linzen of Johns Hopkins, will explore the degree to which explicit innate biases are needed to learn linguistic mappings, whether between linguistic forms (e.g., active/passive or declarative/interrogative) or between forms and meanings. The work will explore a range of neural network models and compare their behavior with that of human language learners.