Algorithm for predicting mathematical formulae from linear strings for mathematical inputs

Abstract

Recently, computer-aided assessment (CAA) systems have been used for mathematics education, with some CAA systems capable of assessing learners’ answers using mathematical expressions. However, the standard input method for mathematics education systems is cumbersome for novice learners. In 2011, we proposed a new mathematical input method that allowed users to input mathematical expressions through an interactive conversion of mathematical expressions from colloquial-style linear strings in WYSIWYG. In this study, we propose a predictive algorithm to improve the input efficiency of this conversion process by using machine learning to determine the score parameters with a structured perceptron similar to natural language processing. In our experimental evaluation, with a training dataset comprising 700 formulae, the prediction accuracy was 96.2% for the top ten ranking by stable score parameter learning; this accuracy is sufficient for a mathematical input interface system.