Learning to Solve First-grade Math Word Problems

Abstract

Math Word Problems is one of the challenges in the Natural Language Processing field. The problem asks the solver to translate from an ambiguous language like the English language to precise language like a mathematical equation. The solver must embed a common background knowledge to be able to understand the context of the question and translate it. In recent years, due to the breakthrough of the Deep Learning and Sequence-to-Sequence model, most researches have been adopting them. They are aiming to model the solver that can solve all the problems. Alternatively, this research aims to understand the relation between the problem and its corresponding equation. What level of knowledge the solver is needed to solve the problem set designed for the first-grade students.