Learning to Solve Arithmetic Word Problems with Verb Categorization

Mohammad Javad Hosseini,Nate Kushman,Oren Etzioni,Hannaneh Hajishirzi

doi:10.3115/v1/d14-1058

Abstract

This paper presents a novel approach to learning to solve simple arithmetic word problems. Our system, ARIS, analyzes each of the sentences in the problem statement to identify the relevant variables and their values. ARIS then maps this information into an equation that represents the problem, and enables its (trivial) solution as shown in Figure 1. The paper analyzes the arithmetic-word problems “genre”, identifying seven categories of verbs used in such problems. ARIS learns to categorize verbs with 81.2% accuracy, and is able to solve 77.7% of the problems in a corpus of standard primary school test questions. We report the first learning results on this task without reliance on predefined templates and make our data publicly available. 1

Highlights

Designing algorithms to automatically solve math and science problems is a long-standing AI challenge (Feigenbaum and Feldman, 1963)
For natural language processing (NLP), mathematical word problems are attractive because the text is concise and relatively straightforward, while the semantics reduces to simple equations
Our contributions are three-fold: (a) We present ARIS, a novel, fully automated method that learns to solve arithmetic word problems; (b) We introduce a method to automatically categorize verbs for sentences from simple, easy-to-obtain training data; our results refine verb senses in WordNet (Miller, 1995) for arithmetic word problems; (c) We introduce a corpus of arithmetic word problems, and report on a series of experiments showing high efficacy in solving addition and subtraction problems based on verb categorization

Summary

Introduction

Designing algorithms to automatically solve math and science problems is a long-standing AI challenge (Feigenbaum and Feldman, 1963). For NLP, mathematical word problems are attractive because the text is concise and relatively straightforward, while the semantics reduces to simple equations. Arithmetic word problems begin by describing a partial world state, followed by simple updates or elaborations and end with a quantitative question. The language understanding part is trivial, but the reasoning may be challenging; for our system, the opposite is true. She gave some of her kittens to Joan. Liz has 5 kittens left and 3 have spots.

Methods

Results

Conclusion