Learning to Automatically Solve Algebraic Word Problems

We introduce an approach that automatically learns to solve algebra word problems. Our algorithm, a pioneering result for this task, reasons across sentence boundaries. It constructs and solves a system of linear equations while simultaneously recovering an alignment of the variables and numbers in these equations to the problem text. The learning algorithm incorporates varied supervision, utilizing either full equations or just the final answers.To solve these problems, our approach necessitates reasoning across sentence boundaries. It aims to identify a system of equations that concisely models the semantic relationships described in the problem text. This work represents, to the best of our knowledge, the first successful learning result for this challenging task.