Knowledge Fusion and Evaluation System with Fusion-Knowledge Measure

Abstract

Knowledge fusion is to combine knowledge from multiple, distributed, heterogeneous knowledge sources. To control the scale of fusion-knowledge and avoid the illogic knowledge, relationship strength between the knowledge elements and semantic relevancy based on ontology were firstly analyzed. Next, Semantic entropy of fusion-knowledge was defined and analyzed using maximum entropy models. On the basis of the analysis above, fusion-knowledge measure with the attribute value weight was formulated to provide guidance for knowledge fusion algorithm. Then, the accelerating fitness function was formulated through fusion-knowledge measure and the genetic simulated annealing algorithm was applied to knowledge fusion by improving population selection and genetic operations. According to the analysis of the balance equation for knowledge entropy, the evaluation mechanism based on information diffusion theory was applied to improving the accuracy of fusion-knowledge. Finally, their effectiveness was demonstrated by an illustrative example. The results show that knowledge fusion algorithm and its evaluation mechanism are beneficial to find the better and potential knowledge, and the semantic relativity and the accuracy of fusion-knowledge are improved.