Research on automatic construction of the dynamic Chinese porn feature knowledge system

As the Web continues to grow, the pornographic texts in varied forms run rampant on Internet, despite repeated prohibitionsm. It severely does harms to the development of people´s mental health and the stability of sociality. In this paper, We propose a novel and effective feature weight-value calculating method named FEFP for such objectionable information, and design a scheme to auto-build system of porn feature knowledge which can be dynamically updated. The experiments are designed and conducted to compare the FEFP with four famous existing methods of Web page feature extraction. Experimental results show that the proposed approach has both higher feature coverage and lower redundancy rates for either porn on conventional forms or porn whose form has been changed. It is very good and very suitable for processing the harmful information in current form which is volatile and increasingly hidden.