Personally, I think public opinion monitoring can be divided into four links: discovery, tracking, handling and reporting, aiming at the existing general network carriers, such as search engines, forums, communities, post bars and Weibo. From the complexity of the carrier and the span of technical environment, we can imagine the technical difficulty of public opinion monitoring automation products. Personally, I feel that the public opinion tools on the market are basically limited and can only be used as auxiliary tools for public relations media. Especially in the traceability of a public opinion event, it is difficult to be accurate, and basically after the outbreak of public opinion events, the spread speed of the Internet age has basically increased geometrically within 4-5 hours. It is even more impossible to judge whether and when an event may break out through the system, so the existing basic public opinion products will inevitably provide some manual services. For business owners or government agencies, the biggest demand is of course to find and control the development of public opinion in time. This is also the main reason why public opinion monitoring products attract customers' attention with so-called whole network monitoring and real-time monitoring. In fact, in view of the existing so-called whole network monitoring, I think it is more effective to monitor public opinion in several new media forms, such as Weibo and WeChat. In particular, Weibo, whether Sina Weibo or Tencent Weibo, as the core features of Weibo, lies in the conciseness of text content, the prominence of keywords and the simplicity of semantic judgment. At the same time, Weibo is the most prone to the outbreak and widespread dissemination of online public opinion. At present, most mainstream media and social celebrities have posted Weibo, so it is the easiest to find public opinion in time through public opinion monitoring in Weibo. Combined with the open platform resources in Weibo, it is easier to analyze the spread of tracking events and find key people in time. Now many companies are doing this part of the application and system research, and have achieved certain results. If we can analyze the propagation path of public opinion Weibo and judge the similarity of Weibo's text through semantic judgment, we can trace it back to the original Weibo of the event. I think it will be more effective and timely than most monitoring systems on the market at present. Combined with timely early warning channels and a period of data analysis and mining, users can form a relatively perfect perception of events. Of course, the above analysis is only aimed at the system level. I think public opinion monitoring will still need manual intervention for a long time to come. Public relations personnel with strong sensitivity and rich experience in handling public opinion events, if they are internet-wide monitoring tools (posting, forums, news, self-media, etc. Of course, it also includes Weibo WeChat), and the multi-user public opinion data analysis station system is still good, which is much more reliable than a system that throws out a lot of money. However, it is obviously more difficult for most business owners and government agencies. Therefore, for government agencies and enterprises, in addition to a relatively reliable public opinion monitoring system, it is more important to try to learn and accumulate online public opinion events that happen all the time, so that they can respond in time when they encounter actual situations.