Abstract: 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.