中文摘要：

为发现语义Web使用记录中所蕴含的有效信息，本文提出了一种挖掘日志本体频繁Web访问模式的方法．该方法引入应用访问规则集和观察集分别表示日志信息动态变化的语义规则和使用事实，并在DL安全的限定下将日志本体和应用访问规则集相结合构成一个推理过程可判定的混合知识库．在此基础上，利用日志本体中事件整分关系的语义构建访问模式学习的事务模型，并采用ILP的方法学习生成频繁用户访问模式树，解决了推理访问模式中非描述逻辑原子的问题．实验结果表明该方法的可用性和有效性．

英文摘要：

In order to discover the useful information from semantic Web usage records, we present an approach for mining the frequent Web access pattems from log ontologies. This method adopts application access-rules to represent the dynamic semantics rules of user-access and adopts observations to represent the usage facts. With the restriction of DL-safety, it combines log ontologies and application user-access rules into a decidable hybrid knowledge base. The transaction mode of access-pattern learning can be extracted form the semantics of the part-whole relations between events in log ontologies. A frequent Web access-pattern tree can be generated by an ILP method from the hybrid knowledge base. This method also solves the problem of reasoning the Web access - patterns with non-DL atoms. The experimental results show that this method is effective and it is quite feasible to solve practical problems.