中圖分類號： TP391.1
文獻標識碼： A
DOI： 10.19358/j.issn.2096-5133.2020.07.009
引用格式： 吳習沫，朱廣宇，張雷. 安全類文章的多文本分類系統的設計與實現[J].信息技術與網絡安全，2020，
39(7)：52-56，60.

Design and implementation of multi-text classification system for security articles

Abstract： At present, the classification labels of security website information are different, and there is no unified classification standard, so that security websites cannot accurately display specific types of security information to users. Faced with a large number of technical article information of security websites, users need to spend a lot of time to identify text categories. So, it′s significant to design a multi-text classification system to advance the user experience and make use of security websites′ efficiency. This paper develops a security text multi-text classification system based on a hybrid model of CNN and LSTM. Based on the Scrapy framework, a web crawler, which supports both customized configuration to extract page data in different layouts and data persistence storage, is used in this system. Based on the mixed model of CNN and LSTM, a multi-text automatic labeling module is designed and implemented to realize the automatic classification of website security information. The rate of classification accuracy has increased by 1.79% and 1.54% in comparison with the traditional CNN and LSTM models respectively. Meanwhile,the F1 value has increased by 1.02% and 0.32%.

Key words : in-depth learning；text categorization；crawler；system