基于领域本体的细粒度用户兴趣建模研究

Toward Fine-grained User Preference Modeling Based on Domain Ontology

用户兴趣模型的建立和维护是个性化推荐服务系统研究和开发中的一个关键问题。细粒度兴趣通过在用户兴趣特征集中区分用户的不同兴趣主题类别来发现,是对粗粒度用户兴趣的进一步挖掘和划分。本文针对以往粗粒度用户兴趣建模存在的不足,对细粒度用户兴趣建模的概念和主要方法进行简介和评述。在此基础上,提出一种基于领域本体和近邻概念聚集的细粒度用户兴趣建模方法（FUPMo）。该方法通过滑动窗口控制用户兴趣概念的计算规模,通过时间窗口和兴趣衰减函数反映用户兴趣的变化。基于军用飞机领域本体OntoAvion和小规模文档库进行的实验研究表明,该细粒度用户兴趣建模方法在应用上具备有效性。

The construction and maintenance of user preference model is a key problem during researching and developing personalized recommendation service system.Fine-grained user preference is the mining and dividing result of user rough-grained preference,and it can be realized by dividing user preference character sets into several different topic classifications.According to the problems of rough-grained user preference modeling methods,a brief statement on concepts and methods of fine-grained ueser preference modeling has been given.Beyond of those,a new fine-grained user preference modeling method based on domain ontology（FUPMo） is proposed,which congregates near neighbored concepts from rough-grained user preference vector to find and describe user＇s multi-aspect interested contents.The FUPMo method uses a slipping-window method to control compute scale,and can reflect user＇s preference variation by utilizing time-window and preference attenuation function.By applying domain ontology OntoAvion and relative domain documents collection,experiments showed that our fine-grained preference modeling method can catch the composition and variation of one user＇s multi-aspect preferences in a reasonable manner.