Discrete Word Embedding for Logical Natural Language Understanding
This work addresses the challenge of integrating natural language understanding with symbolic reasoning for AI planning systems, though it appears incremental as it builds on existing discrete embedding methods.
The authors tackled the problem of making word embeddings compatible with symbolic planning by proposing an unsupervised neural model that learns binary embeddings supporting vector arithmetic, representing each word as propositional statements for classical planning solvers.
We propose an unsupervised neural model for learning a discrete embedding of words. Unlike existing discrete embeddings, our binary embedding supports vector arithmetic operations similar to continuous embeddings. Our embedding represents each word as a set of propositional statements describing a transition rule in classical/STRIPS planning formalism. This makes the embedding directly compatible with symbolic, state of the art classical planning solvers.