基于在线评论信息挖掘的动态用户偏好模型构建

摘要

Purpose/Significance] The paper made up the deficiency of user preference information collection based on rating informa-tion through the advantages of domain ontology in mining information hierarchically and building user preference model. [ Method/Process] Firstly, based on the hierarchy of domain ontology, the hierarchical product attributes set were built through the conceptual map-ping of ontology. Secondly, based on cascading CRFs model and emotional dictionary, the hierarchical analysis of emotional orientation was achieved. And on this basis, the hierarchical mining of user preference information was realized. Finally, the paper built the dynamic user preference model by ontology modeling method, and kept the ontology of user preference dynamic update. [ Result/Conclusion] Simulation experiments conducted simulate an online shopping recommender system which contains 100 users. Through comparative analy-sis with other user preference modeling methods, the accuracy and validity of the model are verified.%[目的/意义]基于领域本体在信息层次化挖掘以及用户偏好模型构建方面的优势，以弥补传统的基于评分信息的用户偏好信息采集的不足。[方法/过程]首先，基于领域本体的层次结构，通过本体概念的映射，构建层次化产品属性集；其次，基于层叠CRFs模型以及情感词典，实现在线评论情感倾向性的层次化分析，并在此基础上，实现了用户偏好信息的层次化挖掘；最后，构建了基于本体建模方法的动态用户偏好模型，以保证用户偏好本体的动态更新。[结果/结论]仿真实验模拟了包含100个用户的网络购物推荐系统，通过与其他用户偏好建模方法的对比分析，该模型的准确性与有效性得到了验证。