Robust semantic systematicity and distributed representations in a connectionist model of sentence comprehension
Stefan Leo Frank, Pim Haselager · eScholarship (California Digital Library) · 2006
There is considerable debate about the amount and kind of systematicity displayed by neural networks.One important requirement is that the systematicity is robust and not overly dependent on specific parameter settings and training details.We present a recurrent neural network model of sentence comprehension that realizes robust semantic systematicity on the basis of distributed representations.