A Study On The Recognition Of Clickbait Online News Based On BERT And Machine Learning Algorithm

With the development of user-operated media in news industry,clickbaits have emerged one after another.The increasing number of clickbaits in recent years has made readers annoyed.Therefore,how to recognize and detect clickbaits has been brought into focus.The traditional method of detecting clickbaits relies on feature engineering and can hardly achieve expected results.The present study,based on the narrative strategies of clickbaits,text analysis,word embedding and machine learning method,use web spider to get data from news website and establishes a model to describe the process of recognizing clickbaits.In this model,both headline key words and the relevance of headlines to text meanings are taken into consideration by word2 vec model and we add three extra features to assist recognize clickbaits.I attempt to use BERT model which is up-to-date to recognize clickbaits.BERT model uses a more advanced language model so that the word embedding result is better.The results shows that BERT model is more effective than traditional models and the accuracy is over 80%.Finally this message build a mature model based on machine learning to recognize clickbaits.