Research On Hot Spot Event Recognition Technology Based On Feature Recognition And Empirical Analysis

Internet hotspot events reflect social dynamics and the will of people's livelihood and have received widespread attention from all walks of life.However,there is a large amount of redundant information in the news portal.It is difficult for people to effectively obtain information related to network hotspot events through manual methods,and some network hotspot events have exploded rapidly and have a huge impact.People need to understand the development trend of the event in time.Therefore,people need computers to automatically discover network hotspot events in a large amount of news information and even identify hotspot events in the early period.Network hotspot event recognition technology has received extensive attention from researchers.Many researchers have used different text representation models to improve the effect of hot spot event recognition and predict the heat of new events based on the similarities between the existing events and the new events.However,in the field of news,there are few studies on improving the efficiency of network hotspot event recognition and there is still a lack of research on early recognition methods of network hotspot events.This dissertation empirically analyzes the characteristics of network hotspot events and early characteristics of network events,improves the text representation method based on the characteristics of news reports to improve the recognition efficiency of network hot events,and proposes an early recognition method of network hot events based on the early characteristics of network events.The research contents and contributions of this article are as follows:(1)This dissertation proposes a KSSP network hot spot event recognition method.This dissertation empirically analyzed the characteristics of news reports of network hot events.It is found that there is more than one topic in the discussion of network hot spot event.In the process of topic shifting,some texts of new topics will briefly elaborate other topics at the beginning of the texts.According to this feature,this dissertation innovatively uses keyword sets and related topic word sets to convert new text expression,which reduces the complexity of news text representation.According to the new text representation method,the Single-Pass algorithm is improved,and a KSSP network hotspot event recognition method is proposed.Experimental results show that the KSSP network hotspot event recognition method proposed in this dissertation not only ensures the recognition effect of network hotspot events,but also improves the recognition efficiency and has better stability.(2)This dissertation proposed a method for identifying hot events in the early period of events based on the early characteristics of network events.Firstly,this dissertation empirically studies the life cycle characteristics of network hotspot events,and rationally defines the early period of network events according to the degree of public opinion influence.Secondly,the empirical study of the performance characteristics of news media and Internet users on network hot events,extracted the early characteristics of 9 network events,including the number of news,comments,media attention,etc.and proposed the method for identifying hot events in the early period of events based on the early characteristics of network events.Finally,Finally,the effectiveness of the method is proved through experiments.(3)Combined with the identification method of network hotspot events studied in this dissertation,I developed a public opinion customization system based on large-scale news data.The network hotspot event recognition method studied in this dissertation is the key to the development of the system.The network hotspot events identified by the system provide data support for the functions of sentiment analysis,event heat cycle analysis and correlation identification.The KSSP network hotspot event recognition method proposed in this dissertation improves the recognition efficiency while ensuring the recognition effect and has certain stability.At the same time,this dissertation extracts the early characteristics of network events in the news field and proposes a new method that can effectively identify hot events in the early period of the event.Therefore,the research in this dissertation can provide theoretical and technical support for the discovery and tracking of hot events in the field of news.