Exploring the Discourse Power Structure in Modern News Communication Based on Text Mining Techniques

In the era of social media, the emergence of a large number of new media has changed the communication pattern dominated by traditional media, and at the same time, it has also changed the discourse power structure of news communication. In this paper, we use the TF-IDF optimization algorithm for keyword extraction and classification of news text, the optimized LDA topic model to mine the topics of news text, and the sentiment analysis of the text, to realize the exploration of the influence mechanism targeting the government’s online discourse power. Public opinion events that are directly related to the government account for only 6.30% of the events of public concern, but the public and media tend to prefer negative government events. Public opinion events in the three areas with a greater connection to the public sector show a strong “broken window effect”, i.e., after the occurrence of a broken window event, the attention to similar events increases sharply, and the aggregation situation is obvious. In the typical cases of negative government public opinion events, positive, neutral, and negative emotions accounted for 21.68%, 14.71%, and 63.61% of the public comments respectively, and the comments on the public opinion events tended to be negative. The government should take timely and positive remedial measures for negative public opinion events to strengthen its discourse power. This paper provides theoretical references for the reconstruction of discourse power in modern news communication.