【摘要】 微博在气象预警信息传播及气象部门收集社会公众信息反馈中发挥了积极作用，并产生了海量级的交互信息，客观分析此类信息对于气象部门及时引领舆论导向和调整发布策略具有较高的参考价值。文中基于2017年1月1日—2020年7月3日大连市重大气象灾害过程，运用自然语言处理、情感分析等方法对大连气象微博中与气象预警相关的评论内容进行情感分析研究。结果表明：公众关注最多的是天气预报准确率、天气过程对于生活影响以及信息发布的及时性。公众对重大天气过程特别是致灾天气过程本身负面评价多于正面，发布策略应根据气象灾害持续时间及时调整预警信息内容和发布频次；公众对大风、冰雹等气象灾害预警信号关注度随季节变化明显，发布策略应针对旅游业、养殖户、种植户等提出专业防范对策；针对发布频率高的大风预警信号，有必要调整业务发布规范，同时关注灾后事件信息的跟踪发布；研究发现，负面情感主要针对天气预报的准确度。更多还原

【Abstract】 Microblog plays an active role in the dissemination of meteorological early warning information and the collection of public information feedback by meteorological departments, and produces a large amount of interactive information. The objective analysis of such information provides a high reference value for meteorological departments to guide public opinion and adjust release strategies in time. Based on the processes of major meteorological disasters in Dalian from January 1, 2017 to July 3, 2020, the comments related to meteorological early warning in Dalian meteorological microblog were discussed using natural language processing, emotion analysis and other methods. The results showed that the public paid most attention to the accuracy of weather forecast, the impact of weather process on life and the timeliness of information distribution, but more negative comments from public on major weather processes, especially the disaster causing weather process. The content and frequency of early warning information should be timely adjusted according to the duration of meteorological disasters. The public’s attention to the early warning signals of meteorological disasters such as gale and hail varies significantly with seasons. The strategy distribution should put forward professional preventive countermeasures for tourism, farmers and planters; For the high frequency gale warning signals, it is necessary to adjust the business distributing specification and pay attention to the tracking and release of post disaster event information. The negative emotions are mainly related to the accuracy of weather forecast.更多还原