Improved Synthetic Weighted Algorithm of Ontology-Based Semantic Similarity Computation

To solve the problems of incomplete consideration and low precision in existing domain ontology semantic similarity computation, an improved synthetic weighted algorithm of ontology-based semantic similarity computation is proposed, mixing path coincidence degree, the shortest distance, and concept property methods. First, the depths of lowest common ancestor (LCA) and an ontology tree are added to the formula of path coincidence degree for distinguishing the influence of LCA depth on similarity when multiple inheritances occur. Second, the analysis of similarity algorithm based on the shortest distance cannot distinguish two situations with the same path distance. One is when the density of LCA is different. The other is a depth difference in the concept pair. So, the number of direct subnodes of the LCA and the depth difference are added to the formula of the shortest distance. Meanwhile, the switch of density factor is set to ensure similarity calculation results between [0,1]. Then, a synthetic weighted algorithm of semantic similarity computation is constructed using the weighting path coincidence degree, the shortest distance, and the concept property. Finally, this algorithm and the other three algorithms in the literature are used to calculate semantic similarity in tea ontology. The results show that this algorithm is closest to expert experience.