喻 鑫，張 亮，冉欽朋，閆龍增

　　(長江水利委員會水文局 長江上游水文水資源勘測局，重慶 江北 400020)

　　摘 要：目前對防洪重大事件新聞輿情研究較少，針對今日頭條平臺中關于長江5號洪水過境重慶期間，頭條新聞的新聞報道和公眾評論組成的新聞輿情進行研究。基于自然語言處理的中文分詞、LDA主題模型等方法對由新聞報道和公眾評論建立的語料進行探究，通過添加防洪先驗知識改進LDA主題模型挖掘效果，從起始、集中、衰退3個階段進行輿情發展思考，從新聞媒體和公眾2個方面對洪水過境重慶的熱度和主題2個維度組成的熱點進行分析，提出防洪重大事件新聞輿情挖掘分析框架。分析結果表明：新聞輿情時間性強，熱度、主題與時間關聯度高，新聞輿情熱點演化過程與洪水演進過程基本一致，可為政府部門把握輿情演化路徑、提前做好輿情預警、掌握輿論主動權提供參考意義。

　　關鍵詞：防洪重大事件;新聞輿情;自然語言處理;長江5號洪水;主題

　　Analysis on news and public opinions on major flood control events based on natural language processing

　　—taking the Yangtze River Flood No. 5 as an example

　　YU Xin，ZHANG Liang，RAN Qinpeng，YAN Longzeng

　　( Upper Changjiang River Bureau of Hydrological and Water Resources Survey，Bureau of Hydrology，Changjiang Water Resources Commission，Chongqing 400020，China)

　　Abstract：So far, there has been little research on the news and public opinions on major flood control events. This paper focuses on the news and public opinions from headlines and public comments on Yangtze River Flood No. 5 in Chongqing on Chinese news platformToutiao. The corpus established by news reports and public comments is analyzed based on natural language processing method such as Chinese word segmentation, LDA topic model etc. The mining performance of LDA topic model is improved by adding priori knowledge of flood control. This paper considers the development of public opinions from three stages (i.e., initiation, concentration and recession) to analyze the hot spots in two dimensions (i.e., popularity and topics) of the flood event in Chongqing from two aspects：news media and the public. The mining and analysis framework of news and public opinions on flood control events is then proposed. The results indicate that the news and public opinions has strong timeliness, showing high correlation between popularity, theme and time. The evolution process of hotspots of news and public opinions is basically the same as that of flood evolution, which can provide reference for government departments to grasp the evolution path of public opinions, develop early warnings of public opinions, and lead the initiative of public opinions.

　　Key words：major flood control events;news and public opinions;natural language processing;Yangtze River Flood No. 5;topics