基于生成式对抗网络的互联网评论区舆情检测方法

Public Opinion Detection Method in Internet Comment Area Based on Generative Adversarial Network

互联网评论区的数据量庞大,且用户评论的更新速度极快,需要处理的数据量巨大,导致检测互联网评论区舆情的速度较慢,为此研究基于生成式对抗网络(Generative Adversarial Networks,GAN)的互联网评论区舆情检测方法。首先预处理信息,包括去除噪声和标准化文本数据。其次,利用GAN生成模拟的舆情信息,增强模型的泛化能力。再次,提取舆情信息的特征,反映文本的情感倾向和主题内容。最后,通过主题分类算法实现舆情检测,准确判断评论的主题类别和情感态度,为舆情监控和应对提供有力支持。实验结果表明,即使评论数量高达492.6万条,该方法依然能够保持每秒处理475000条评论的速度,具有显著的处理速度优势。

The amount of data in the Internet comment area is huge,and the update speed of user comments is extremely fast.The amount of data that needs to be processed is huge,which leads to the slow speed of detecting public opinion in the Internet comment area.Therefore,we study the public opinion detection method in the Internet comment area based on the Generative Adversarial Networks(GAN).Firstly,preprocess the information,including removing noise and standardizing text data.Secondly,using GAN to generate simulated public opinion information enhances the model’s generalization ability.Once again,extract the features of public opinion information to reflect the emotional orientation and thematic content of the text.Finally,public opinion detection is achieved through topic classification algorithms,accurately determining the topic category and emotional attitude of comments,providing strong support for public opinion monitoring and response.The experimental results show that even with a high number of 4.926 million comments,this method can still maintain a processing speed of 475000 comments per second,with significant processing speed advantages.