Efficient n-Gram-Based String Matching in Electronic Testing at Programming

Abstract

AbstractThe purpose of this study is to explore the grammatical properties and features of string matching based on n-grams techniques and to apply them to electronic testing at programming languages. Because of the intensive and extensive availability of internet in the academic environment, there exists a real need for computational procedures within artificial intelligence methods that support assessment of examination questions for uniformity and consistency. There are many computer-aided assessment packages, mainly designed for single and multiple choice tests, but they are not suitable for electronic testing at programming languages. n-grams based string matching is being successfully applied to document retrieval and other natural language processing applications. Generalized n-grams matching during substring processing tends to be time-consuming since there are N 2 n-grams extracted where n is the length of a (sub)string. The choice of selecting parameter n in n-grams approximately is an important task since the large size of n leads to polynomial growth and its small size may shorten search time significantly. As the result, some new string matching methods based on n-grams are proposed for the improvement of generalized n-grams. Experiments are conducted with the method using programming language codes as both pattern and text matching. The results are compared to chosen existing methods. We found the obtained results very promising and suggest that the proposed methods can be successfully applied to electronic testing at programming languages as an intelligent support for teachers involved in e-learning processes.