A Method of Harmful Information Identification and Web Page Classification Based on Multiple Instance Learning

Abstract

The invention discloses a method for classifying webpages based on multi-instance learning. The method includes: designing a relative size sorting forward comparison method to extract effective images in webpages, and extracting relevant texts of effective images according to the tree structure of webpages; An effective image and its related text are taken as an example in the web package, and the image bag-of-words model and the text bag-of-words model are used to generate a description of the effective image and its related text, and the two are combined as a description of the example; using multiple examples Check out the drug pages for classification. The method of the present invention uses the images and related texts contained in the webpage as examples in the webpage package, so that the algorithm is more in line with the actual distribution of the content of the webpage, and can make full use of the effective information of the webpage, and dig deep into the image information and text information Complementarity, and ultimately achieve better results than only using unimodal information for classification.

Description

technical field [0001] The invention relates to the field of network content security, in particular to a method for identifying harmful information and classifying webpages based on multi-instance learning. Background technique [0002] While the Internet promotes social progress and development, it also provides great convenience for the dissemination of various harmful information. Such harmful information increasingly endangers normal social activities and healthy value systems, and is especially detrimental to the healthy growth of young people. Maximizing the positive effects of the Internet and suppressing or eliminating its negative effects will help purify the Internet environment, promote social progress, and care for the healthy growth of young people. Harmful information on the Internet includes pornography, drugs, violence, terror, reactionary, etc. The harm of drug information is more than that of other harmful information. [0003] In the Internet, web pages...

Images