节点文献
基于用户兴趣模型的新闻自动推荐系统
【作者】 李伟;
【导师】 钱松荣;
【作者基本信息】 复旦大学 , 通信与信息系统, 2009, 硕士
【摘要】 个性化服务就是为不同用户按照其需求提供不同的服务,根据用户的特征信息建立不同的服务策略和服务模式。个性化服务通过收集和分析用户信息来学习用户的兴趣和行为,从而实现主动推荐的目的。本文设计并讨论了在构建用户兴趣模型的基础上,主动向用户推荐符合用户新闻偏好的新闻自动推荐系统。本系统能够通过对用户阅读习惯和阅读行为的分析,建立了用户阅读新闻的兴趣模型。根据用户兴趣模型,系统过滤出符合用户兴趣模型的新闻,主动推荐给用户。基于用户兴趣模型的新闻自动推荐系统可以极大的方便用户阅读新闻的过程,降低用户对海量新闻进行过滤的负担。本文研究重点包括:一、如何建立和表达用户的兴趣模型研究了当前主流的兴趣模型建立方法,根据新闻领域的特点,提出了以新闻分类偏好表的形式表达用户在新闻领域兴趣模型的方法。二、如何判断一篇新闻文本符合用户的兴趣模型系统成功运用了基于概率分析的文本自动分类算法。这个算法建立了文本关键词和新闻分类之间的关系,计算出分析出文本属于不同分类的概率,即文本分类矢量。根据新闻文本的分类矢量和用户的兴趣模型,系统判断该新闻是否符合用户的兴趣模型。三、如何对兴趣模型进行更新系统充分实现了基于用户反馈的兴趣模型更新过程,并对兴趣模型的增强和减弱过程,进行了积极的尝试。
【Abstract】 Personalized service provides different services to its users, according to the characteristics of their information and creates different service strategies and service models. Personalized service collects and analyzes user information to learn their interest and behavior in order to achieve the purpose of the recommended initiatives. This article discusses the design of automatically news recommendation system and builds a model for user interest comply with user preferences.The system can build interest model for users based on their reading habits and behavior analysis. Based on user interest model, the system filters out the news, takes the initiative to recommend them to the user. Model based on user interest can automatically provide a great user-friendly process of reading the news, and reduce the burden of filtering news information.Areas of Research Interests:1. How to create and express user’s interest modelThe article analyzes the main interest model in use, and proposes to express user interest by news preferences, in accordance with the characteristics of the news information.2. How to determine the text of news complies with user interest model Successfully applying text categorization algorithm based on probability analysis.This algorithm establishes the relationship between the text keywords and the belonging categories, namely, text classification vector. According to text classification vector and the user’s interest model, system determines whether the news information in line with user interest.3. How to update user interest modelThe system realizes the process of updating user interest model based on user feedback. The process may strengthen or weaken user interest in the model.
【Key words】 interest model; news recommendation; text keyword; text classification; interest model update;