结合用户交互行为和资源内容的资源推荐

Providing Resource Recommendation Based on Interactive Behavior and Resource Content

为了更好地管理任务以及与任务相关的资源,使用户集中注意力在任务本身上,减少用户的交互负担,提出一种基于隐式Dirichlet分配(LDA)模型的任务建模方法.通过将用户的交互行为按时间片进行切分,实现了时间片序列—任务—文件与LDA模型中的文章—主题—单词的对应,经过LDA方法的学习,得到了时间片—任务的概率分布和任务—文件的概率分布;为了对任务模型进行补充,进一步提出了基于资源内容的主题分析方法,并用LDA方法建立了主题模型;最后通过对资源的关联关系分析,实现了一个结合任务模型和主题模型的资源推荐系统.实验结果表明,任务模型能够有效地发现用户的主要任务和主要文件.

To improve management of tasks and related resources, help users increase their concentration on tasks and reduce their interaction burden, a new task modeling method based on Latent Dirichlet Allocation （LDA） model is proposed. By segmenting the user＇s semantic behavior according to time slices, the mapping of time slice--task--file in user activity to document--topic-- word in the LDA model is realized. After the learning process of LDA method, the probability distribution of time slice to task and the probability distribution of task to file are attained. In order to complement the task model, a topic analysis method based on the content of resources is proposed, and the topic model is built using LDA method. Finally, the relevance of resources is analyzed and a resource recommendation system based on task model and topic model is built. Experimental results show that the task model can find the user＇s main tasks and main documents effectively.