Research On The Method Of Network Event Heat Analysis Based On Semantic Fingerprint Correlation

With the rapid development of the Internet, the Internet media came into being, the traditional media companies also established the news website. The characteristics such as quick, interactive and open of network communication make people pay close attention to the hot spots on the Internet, and express their views and participate in the discussion. With people's attention and negative hot event's damage may cause bad influence on social harmony, therefore, in the beginning of the hot events timely detection of hot events and take appropriate measures to deal with, is the furher research direction of hot events.The current hot event detection method is mainly classified by the method of text processing to classify the information on the network, and then the heat analysis of each class of events. This type of method is an ex post analysis, if it can be found in the embryonic stage of the event and the timely treatment of the event will be able to effectively reduce the impact So, the goal of this study is to analyze the heat of the event in the embryonic stage of the event, and to judge whether it will become a hot spot.When a new event has just produced, the number of reports, comments, click on the number of the few, through the heat calculation formula can't determine whether it will become a hot event. Analysis of the characteristics of hot spots, and found that there are a lot of hot events related to the incident occurred in the past. This indicates that the hot events are related to each other, so the heat of the current event can be analyzed by the hot events that have already occurred.In this paper, we first put forward the concept of event semantic fingerprint description of the event, the semantic fingerprint of the event is divided into four parts, the subject, the theme, the time and the place, from the four parts to describe an event. The subject, time and place of the event are named entities, which can be obtained by named entity recognition technology. According to the characteristics of each part, the contribution of the event is calculated. The theme of the event is mainly to describe the content of the event, according to Kolmogorov complexity theory and Co word matrix we can get the most representative event related features of the word as the theme of the event. Event based semantic fingerprints describe an event according to the characteristics of each aspect of the event, which is more perfect than the traditional method of text vector representation.And based event semantic fingerprint calculation of degree of correlation between the two events, the associated event graph is established, the analysis of the heat of the figure in each event, according to the idea of PageRank algorithm get the importance of each event, the product of the event's heat vector and the event's importance vector is calculated, and the heat of the current event is obtained. The correlation degree between the four aspects of the semantic; fingerprint of the event is calculated respectively, the weight matrix is established by taking the weight of the event correlation degree as the edge. Iterative calculation of the importance of events in the event graph, and the importance of each event in the association graph is obtained.Finally, the semantic fingerprint of the event is compared with the TF-IDF, BM25 in the retrieval model, and it is found that the semantic fingerprint of the event can better express an event. The event correlation graph model is calculated to compare the heat of the event and the actual heat. The prediction effect is very good, so the validity of the event correlation method is explained.