Distributional formal semantics

Natural language semantics has recently sought to combine the complementary strengths of formal and distributional approaches meaning. However, given fundamentally different ‘representational currency’ underlying these approaches—models world versus linguistic co-occurrence—their unification proven extremely difficult. Here, we define Distributional Formal Semantics, which integrates distributionality into a semantic system on level models. This approach offers probabilistic, distributed meaning representations that are inherently compositional, naturally capture fundamental notions such as quantification entailment. Furthermore, show how probabilistic nature allows for inference, information-theoretic notion “information” (measured in Entropy Surprisal) follows from it. Finally, illustrate can be derived incrementally input using recurrent neural network model, resultant incremental construction procedure intuitively captures key phenomena, including negation, presupposition, anaphoricity.