Applying Ontology and VSM for Similarity Measure of Test Questions

Abstract

Vector space model (VSM) is a common method for measuring test questions similarity in large-scale item bank system. VSM is limited in accurately representing the knowledge relationship and the potential semantic relations of different characteristic words, hence this paper proposes a method of test questions similarity measure called OVSM-TQSM which combines domain ontology and VSM. OVSM-TQSM can reveal the intrinsic relationship among words by using the constructed domain ontology which integrates with the tree structure and the graphics structure. Incorporated with eigenvectors and the weight of words in VSM, OVSM-TQSM calculates the similarity of test questions. A large number of experimental results demonstrate that the novel approach is feasible and effective. Comparing with the traditional method based on VSM, OVSM-TQSM has the advantages of higher accuracy and little unnecessary laborious pre-processing.