Learning and analyzing vector encoding of symbolic representations

Roland Fernandez, Asli Celikyilmaz, Rishabh Singh, Paul Smolensky

We present a formal language with expressions denoting general symbol structures and queries which access information in those structures. A sequence-to-sequence network processing this language learns to encode symbol structures and query them. The learned representation (approximately) shares a simple linearity property with theoretical techniques for performing this task.

Knowledge Graph

arrow_drop_up

Comments

Sign up or login to leave a comment