Abstract: To solve the problem of inference on mass ontology, a new algorithm was proposed based on cloud computing platform. Ontology schema was transformed into graph, inference strategy was designed accordingly. Inference algorithm was designed based on the computing model of Map/Reduce. After one time iteration, mass ontology could be inferred in the course of Map. Later, duplicated triples were eliminated in the course of Reduce. The experimental results show that inference of one hundred million triples costs less than four minutes. The algorithm is effective.

Key words: cloud computing, mass, ontology, inference

摘要： 为了解决海量本体可推理的问题，以云计算平台为支撑，将本体schema转换为图结构，并设计相应的推理策略，以Map/Reduce为计算模型，重写推理规则，设计推理算法，在Map过程中通过一次迭代实现推理，在Reduce过程中消重，解决了海量资源描述框架模式(RDFs)本体的推理问题，在合理时间实现了海量本体的推理。实验显示，一亿条三元组的推理时间没有超过4min，证明算法是有效的。

关键词: 云计算, 海量, 本体, 推理