Abstract: An ontology-based knowledge fusion framework is presented to improve semantic specification and accuracy of fusion-knowledge.The framework includes the construction of meta-knowledge sets,calculation of fusion-knowledge metric,knowledge fusion algorithm,and post-processing for fusion-knowledge.Knowledge fusion pattern is analyzed and the construction method for the meta-knowledge set is presented according to the relationship between knowledge elements in domain ontology.Semantic relativity is analyzed using maximum entropy models.Fusion-knowledge metric with relationship strength and weight of knowledge elements is formulated.Genetic simulated annealing fusion algorithm,fusion rules and evaluation mechanism based on information diffusion theory are developed.Finally,the effectiveness of the knowledge fusion framework is demonstrated by an illustrative example.The results show that it is beneficial to control the scale of new knowledge,and improve semantic relativity and accuracy of fusion-knowledge.