Abstract: The cores of granular computing are particles, grain layer and the grain structure. The conceptions “ontology grain” and “compatible grain” are defined by application of granular computing ideas and ontology model, the ontology grain set and ontology tree generation algorithms are provided. These algorithms produce initial ontology grain set with compatible class, and extend other ontology grains with the vector of connotation IG, the last, build lattice hierarchy and conception tree model of ontology with the vector of relation RG. The empirical research of traditional Chinese medicine ontology shows that these algorithms are correct and efficient, and provide a good technical way for ontology learning.

Key words: Granular Computing（GrC）, ontology learning, ontology grain, compatible grain, ontology tree

摘要： 粒计算的核心是粒子、粒层和粒结构。应用粒计算思想和本体论模型定义了本体粒和相容粒概念，给出了计算本体粒集和本体树的生成算法。该算法通过相容类产生初始本体粒集，应用本体粒的内涵分量[IG]扩展其他本体粒，最后联合关系分量[RG]和本体粒的关系构建本体粒集的格分层结构，借助加权树思想生成本体的概念树模型。经过中医喘证本体的实证研究，说明算法正确高效，为本体学习提供了很好的技术途径。

关键词: 粒计算, 本体学习, 本体粒, 相容粒, 本体树