基于主题提取和记忆模型的新闻推荐系统设计

互联网的飞速发展产生了＂信息过载＂问题,新闻推荐系统可以帮助用户快速找到符合自己兴趣的新闻。文章完成了一个基于主题推荐的新闻推荐系统的研究和实现,针对该系统的关键部分即新闻主题和用户兴趣模型做了深入的研究并进行了程序上的实现。文章提出了一种基于主题的新闻推荐方式,将概率主题模型引入到文本推荐中,并深入研究了采用Gibbs抽样算法的LDA主题提取模型。文章在研究基于记忆的用户兴趣模型基础上提出了相应的改进方法,主要是针对记忆模型对用户长短期兴趣不适应的问题,提出了根据用户访问频率来确定不同兴趣度衰减速度的改进方案。

News Recommender System Design Based on Subject Extraction and Memory Model

The rapid development of the Internet has created the "information overload" problem.News recommender system can help users quickly find the news of their own interests.This paper has completed the news recommender system research and implementation based on subject recommendation.According to the key part of the system news topics and user interest model,a deep research and the realization of the procedure have been made.This paper proposes a news recommend way based on subject,probabilistic topic model is introduced into the text recommended,The LDA topic extraction model using Gibbs sampling algorithm is in further research.This paper presents the improvement methods based on the study of memory-based model of user interest.For the memory model of users long in adjusting to short-term interest,an improvement plan is proposed which depending on the user access to determine the different interest degrees frequency attenuation speed.