用户自适应的社会标签生成和优化模型研究

项目名称： 用户自适应的社会标签生成和优化模型研究

项目编号： No.61272277

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 自动化技术、计算机技术

项目作者： 朱福喜

作者单位： 武汉大学

项目金额： 81万元

中文摘要： 标签是Web 2.0下用户标注网络信息资源的重要手段，应准确地体现用户对资源的高度理解。本项目旨在研究用户自适应的社会标签生成和优化模型，以便能够对特定网页生成最合适当前用户背景的标签。为此，拟从以下几个方面展开研究：首先，建立一定规模的用户标注语料库，并设计和训练概率生成模型来模拟用户对特定文档的标注过程，以此推导出用户对标签的偏好程度，该模型还将考虑时态因素，以反映用户的实时兴趣；然后，对新网页标注时，借助维基百科和训练集进行对网页扩展，生成标签树，再利用训练好的生成模型获取用户对标签的偏好程度，并综合考虑用户的偏好度、标签之间的冗余度以及主题的覆盖度，采用多目标优化技术产生一组理想的标签；最后，考虑到标签的自适应性，反馈机制也纳入模型研究之中。本项目的研究对于提高Web信息检索质量，捕获Web信息中的热点话题以及应用到计算广告学中提高网页、用户和广告三者的精确匹配都有着重要意义。

中文关键词： 标签提取；标签传播；情感分类；标签推荐；上下文感知推荐

英文摘要： Social tagging, which is an important means of annotating Web information resource in Web 2.0, should accurately reflect users' deeply understanding to the resource. This project aims to study the adaptive model of generation and optimization of social tags, by which the most matching tags for a specific user could be generated with a given webpage. With this aim, we plan to conduct the study from the following aspects: Firstly, a user-annotated corpus with a certain scale should be constructed, and the Probabilistic Generative Model should be designed and trained for simulating the user's annotating process to the given document. In this way, the degree of the user's preference to the tags could be deduced. Besides, the model would be introduced with the temporal factor, which could help tracking and finding the real-time interest of the user. Secondly, when generating tags for the new webpage, we extend the given webpage based on Wikipedia and train set, and then generate a tags tree. And then, we obtain the user's degree of preference using the well-trained generative model. By tags tree and generative model, we generate a set of perfect tags using Multi-Objective Optimization Algorithm. These objectives include three factors: the user's preference of the tags, the redundancy of the tags and the rate of cover

英文关键词： Tag extraction；Label Propagation；Emotional Classification；Tag Recommendation；Context - Aware Recommendatio