Abstract
Geometry problem solving presents a formidable challenge within the NLPcommunity. Existing approaches often rely on models designed for solving mathword problems, neglecting the unique characteristics of geometry math problems.Additionally, the current research predominantly focuses on geometrycalculation problems, while overlooking other essential aspects like proving.In this study, we address these limitations by proposing the Geometry-AwareProblem Solver (GAPS) model. GAPS is specifically designed to generate solutionprograms for geometry math problems of various types with the help of itsunique problem-type classifier. To achieve this, GAPS treats the solutionprogram as a composition of operators and operands, segregating theirgeneration processes. Furthermore, we introduce the geometry elementsenhancement method, which enhances the ability of GAPS to recognize geometryelements accurately. By leveraging these improvements, GAPS showcasesremarkable performance in resolving geometry math problems. Our experimentsconducted on the UniGeo dataset demonstrate the superiority of GAPS over thestate-of-the-art model, Geoformer. Specifically, GAPS achieves an accuracyimprovement of more than 5.3% for calculation tasks and an impressive 41.1% forproving tasks. Notably, GAPS achieves an impressive accuracy of 97.5% onproving problems, representing a significant advancement in solving geometryproving tasks.