Steps toward accurate math word problem translation

Abstract

In this paper, a novel multi-phase architecture for an accurate math word translation system is proposed. This system, "MathMaster", was designed to help students to gradually develop skills required to solve math word problems. Math word problems are difficult for students because they require the student to comprehend the verbal problem, to identify relevant information in a word problem and to translate word problem into equation. MathMaster guides students at their own pace through step-bystep instructions and at a level adapted to each student automatically to the solution. In addition, MathMaster keeps the performance record of each student and provides them feedback at the end of the practice. MathMaster performs the translation in three phases: word problem understanding, information extraction and equation formation. Word problem understanding provides the word meaning requested by a student and help a student to understand the word problem through a question and answer process. Next, information extraction extracts the questions asked in the problems which will be used as the clues to eliminate irrelevant information. Then, equation formation selects the appropriate operators and operands for forming a complete equation. The initial experiment and examples are discussed throughout the paper. ©2009 IEEE.