From Word Models to World Models: Translating from Natural Language to the Probabilistic Language of Thought

Rational meaning construction, a computational framework for language-informed thinking that combines neural language models with probabilistic models for rational inference. Linguistic meaning is framed as a context-sensitive mapping from natural language into a probabilistic language of thought (PLoT)--a general-purpose symbolic substrate for generative world modeling.