Discrete Word Embedding for Logical Natural Language Understanding

TL;DR

This paper introduces an unsupervised neural model that learns binary word embeddings representing propositional statements, enabling logical reasoning and compatibility with classical planning solvers.

Contribution

It presents a novel discrete embedding method that supports vector operations and aligns with symbolic planning formalism, bridging neural and symbolic NLP.

Findings

Binary embeddings support vector arithmetic operations

Embeddings represent words as propositional statements

Compatibility with classical planning solvers achieved

Abstract

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.