An Improvement of Similarity Measurement Based on Semantic Units

Because of some flaws of previous similarity measurement models, we extend the concept attributes and propose a method to measure the similarity based on the semantic unit. We use semantic units to express the concept meanings, and use the supporting degree to show the different contributions of different semantic units. By introducing the psychological knowledge of the relevance, the resemblance, and the non-symmetry of similarity, we measure the similarity of concepts successfully. We treat relation as particular concept, and we find a new method to measure the ontology similarity successfully based on semantic units. Finally, our experiments prove that our results are more reasonable, and it is better than other methods.