A problem solver for quadratic functions

Abstract

Several studies of mathematical and arithmetical problem solvers have been made in an attempt to build systems that perform human intellectual activities. Most of such systems, however, support only simple problems. In order to solve more complicated problems, such as those presented in university admission tests, it is necessary to combine forward reasoning and backward reasoning while using the given constraints and data. In this paper we present MAP-Q, a system for solving problems involving quadratic functions. The MAP-Q system represents quadratic function problems as an extended semantic network and finds solutions by applying rules to propagate constraints and data across the network. Experimental results show that MAP-Q is capable of solving real problems presented in admission tests. Many mathematical problems include multiple questions. In such cases, MAP-Q extracts from the extended semantic networks only those portions that are necessary for a given question. The system separates the solution process for each question, and the results obtained for one question are used for solving the following questions. © 2001 Scripta Technica, Syst Comp Jpn, 32(2): 32–42, 2001