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

Ishadi Jayasinghe, Surangika Ranathunga

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.