基于BERT-LSTM模型对高校网络舆情的实证分析

Empirical analysis of university online public opinion based on the BERT-LSTM model

随着自媒体的迅速发展,网络舆情对高校形象和学生心理造成的影响日益凸显。未经全貌了解的信息可能引发学生的偏见和质疑,甚至损害高校的声誉和爱国主义教育建设。因此,旨在构建一个有监督的语义分析模型,以识别网络上的高校负面舆情信息,便于及时采取措施应对。首先,抓取新浪微博热搜页面数据,进行数据清洗,并利用BERT预训练模型和双向LSTM分类模型进行训练,得到舆情文本分类模型。其次,通过模型评测指标计算,验证得到模型的有效性和稳定性。最后,通过实例证明分类模型在识别和应对高校负面舆情中的实际效果。这一模型有助于校方在舆情发酵前即时感知,并迅速还原真相,稳定校园舆论,维护良好的校园形象和学生心理健康。

With the rapid development of social media,the impact of online public opinion on the image of universities and the psychological well-being of students is becoming increasingly prominent.Misinterpretation of information can lead to biases and doubts among students,potentially damaging the reputation of universities and undermining efforts in patriotic education.Therefore,this study aims to construct a supervised semantic analysis model to identify negative public opinion about universities on the internet,allowing for timely measures to be taken in response.Firstly data is collected by scraping the hot search page of Sina Weibo,followed by data cleaning.Subsequently,a sentiment analysis model is trained using BERT pre-trained models and bidirectional LSTM classification models to classify public opinion text.Secondly,the effectiveness and stability of the model are verified through the calculation of model evaluation metrics.Finally,the practical effectiveness of the classification model in identifying and responding to negative public opinion about universities is demonstrated through several examples.This model aids university administrators in promptly perceiving public opinion trends,swiftly restoring the truth,stabilizing campus discourse,and maintaining a positive image of the university and the psychological well-being of students.