Abstract: Filtering of network sensitive information is an important and complicated task.Aiming at the problems of detection time lag,low accuracy and poor adaptability,we proposed an adaptive multi filtering model of network sensitive information which is based on the research object of Chinese text media from the internet and uses the technologies of opinion mining,machine learning,high performance computing and natural language processing to recognize the sensitive information adaptively from the point of whole and semantic.The research on adaptive multi filtering model of network sensitive information will give some technical supports for the public opinion monitoring,intelligent business and developing systems of aid making decision application.