基于知识和改进深度学习的网络主题文本快速过滤方法  

A Fast Filtering Method for Web Topic Text Based on Knowledge and Improved Deep Learning

作　　者：刘丽娟 Liu Li-juan(Shanghai Branch of National Computer Network Emergency Response Technical Team/Coordination Center of China,Shanghai 201315,China)

机构地区：[1]国家计算机网络应急技术处理协调中心上海分中心,上海201315

出　　处：《科学与信息化》2023年第19期79-81,共3页Technology and Information

摘　　要：现阶段,网络主题文本过滤主要存在知识重用率较低、识别不准不全等问题。为此,本文提出基于知识和改进深度学习的网络主题文本快速过滤方法。该方法借助图谱嵌入内部知识,联系上下文嵌入外部知识,并建立一个改进深度学习模型,改进处是将内、外部知识作为词向量输入,增强语义潜在关联性。结果表明,该方法不仅提升了主题识别准确率F1值,且相比关键词法、互信息法、深度学习法,处理时间大大缩短,提升了文本过滤效率。At the present stage,the main problems of network topic text filtering are low knowledge reuse rate,inaccurate and incomplete recognition.Therefore,this paper proposes a fast filtering method for network topic text based on knowledge and improved deep learning.This method embeds internal knowledge with the help of knowledge graph,embeds external knowledge with the context,and establishes an improved deep learning model,which improves by using internal and external knowledge as word vector input to enhance the potential correlation of semantics.The result shows that this method not only improves the F1 value of topic recognition accuracy,but also greatly shortens the processing time compared with the key word method,mutual information method and deep learning method,in this way,it improves the text filtering efficiency.

关 键 词：深度学习 知识嵌入 语义分析 数据挖掘 人工智能 

分 类 号：G63[文化科学—教育学]