中文摘要：

这篇论文描述一个聚集缓存代替(ACR ) 算法的理论，实现，和试验性的评估。由考虑申请背景，小心地选择重量价值，用计算类似的一个特殊公式，并且为更得到由类似聚类本体论嵌入深关系， ACR 把本体论类似与目标的价值相结合并且决定哪个目标将被代替。Wedemonstrate 通过实验的 ACR 的实用性。聚集树根据申请格不同地完全被创造，这被发现的(a) 。因此，聚类能根据用户感觉更精确地指导内容改编并且能与不同偏爱满足用户。(b) 在把这个新方法与使用得广泛地的 algorithmLast-Recently-Used (LRU ) 和 First-in-First-out (FIFO ) 方法作比较以后， ACR 超过，这被发现以后二在精确性和可用性。(c) 它有更好语义的解释并且使改编成为个性化的更多和更多精确。

英文摘要：

This paper describes the theory, implementation, and experimental evaluation of an Aggregation Cache Replacement （ ACR ） algorithm. By considering application background, carefully choosing weight values, using a special formula to calculate the similarity, and clustering ontologies by similarity for getting more embedded deep relations, ACR combines the ontology similarity with the value of object and decides which object is to be replaced. We demonstrate the usefulness of ACR through experiments. （a） It is found that the aggregation tree is created wholly differently according to the application cases. Therefore, clustering can direct the content adaptation more accurately according to the user perception and can satisfy the user with different preferences. （b） After comparing this new method with widely-used algorithm Last-Recently-Used （LRU） and First-in-First-out （FIFO） method, it is found that ACR outperforms the later two in accuracy and usability. （c） It has a better semantic explanation and makes adaptation more personalized and more precise.