A methodology to evaluate triple confidence and detect incorrect triples in knowledge bases

Abstract

The accuracy of the contents of a knowledge base determines the effectiveness of knowledge service applications, thus, it is necessary to evaluate the confidence of triples when a knowledge base is built. This study introduces a generic computational methodology to compute the confidence values of triples in knowledge bases and detect potentially incorrect ones for further verification. The major contributions of the proposed methodology are as follows: (1) A process to compute the confidence values of triples is designed; (2) New algorithms are proposed to adjust the term frequency and inverse document frequency values of each triple; (3) A method to build a support vector machine (SVM) classifier based on the selected triples used for incorrect triple detection is presented.