Two-Step Memory Networks for Deep Semantic Parsing of Geometry Word Problems

Abstract

Semantic parsing of geometry word problems (GWPs) is the first step towards automated geometry problem solvers. Existing systems for this task heavily depend on language-specific NLP tools, and use hard-coded parsing rules. Moreover, these systems produce a static set of facts and record low precision scores. In this paper, we present the two-step memory network, a novel neural network architecture for deep semantic parsing of GWPs. Our model is language independent and optimized for low-resource domains. Without using any language-specific NLP tools, our system performs as good as existing systems. We also introduce on-demand fact extraction, where a solver can query the model about entities during the solving stage that alleviates the problem of imperfect recalls.