项目名称: 基于社会媒体异质关系挖掘的用户兴趣建模方法研究
项目编号: No.61501463
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 无线电电子学、电信技术
项目作者: 郑楠
作者单位: 中国科学院自动化研究所
项目金额: 19万元
中文摘要: 随着社会媒体广泛渗入用户网络生活,互联网产生了大量内容丰富的异质信息,如形式多样的媒体协同与用户互动。深入分析这类异质信息,进而准确理解用户的意图和兴趣,有效实现个性化信息推荐,成为时下社会媒体应用和发展的关键。传统的用户兴趣建模方法假设个体间只存在单一关系,不利于处理异质数据,本项目旨在通过机器学习和社会网络分析方法,研究基于异质关系挖掘的用户兴趣建模方法。主要研究内容包括:(1)采用主题模型结合语义词典的方式实现文本语义挖掘,以概率的形式将词语转化为用户兴趣层次化主题描述;(2)通过概率模型将用户参与社区的行为映射为用户对主题的喜好程度,实现基于社区的用户兴趣描述;(3)通过社会网络分析方法分析用户好友网络的结构特征,深入分析用户间交互模式,理解用户兴趣倾向;(4)在模型学习的不同阶段融入适当的交互元素,提出改进的潜语义生成模型融合机制,提高协同推荐的准确率和可解释性。
中文关键词: 社会媒体;异质关系;用户兴趣建模;个性化服务
英文摘要: With the rapid development of social media, massive rich media information has been generated, such as complicated information association between users and multimedia content. How to effectively conduct data mining for accurate user modeling and recommendations becomes the key problem to the development of modern Internet. Traditional user interest models focus on learning user interests from single relation, which cannot be applied in social media. This proposal aims to construct user interest models using rich user interactive resources in social media based on machine learning and social network analysis. The key issues includes: (1) Research on text mining based on topic model and semantic dictionary for user modeling; (2) Research on community mining based on probability model and topic model; (3) Research on user network analysis based on social network analysis and structural properties investigation; (4) Designing recommendation models to fuse different elements to improve the accuracy and explanation.
英文关键词: Social Media;Heterogeneous Relation;User Modeling;Personalized Services