Proceedings of the Annual Meeting of the Cognitive Science Society

Increasingly, cognitive scientists have demonstrated interest inapplying tools from deep learning. One use for deep learning isin language acquisition where it is useful to know if a linguisticphenomenon can be learned through domain-general means.To assess whether unsupervised deep learning is appropriate,we first pose a smaller question: Can unsupervised neural net-works apply linguistic rules productively, using them in novelsituations? We draw from the literature on determiner/nounproductivity by training an unsupervised, autoencoder networkmeasuring its ability to combine nouns with determiners. Oursimple autoencoder creates combinations it has not previouslyencountered and produces a degree of overlap matching adults.While this preliminary work does not provide conclusive evi-dence for productivity, it warrants further investigation withmore complex models. Further, this work helps lay the foun-dations for future collaboration between the deep learning andcognitive science communities.