Twitter Event Detection Under Spatio-Temporal Constraints

Billions of data spread on Twitter every day, which carries a lot of information. It is meaningful to mine the useful information and make it valuable. The purpose of Twitter event detection is to detect what happened in our real life from these unstructured data. We introduce the spatio-temporal information of tweets into event detection. The event detection can be divided into three steps in this paper. First, we use the space difference between event words and noise words and introduce the relationship between words, then we can build a model to separate event words and noise words. Then we define the similarity between event tweets from three different aspects, which make up for the shortcomings of existing methods. Finally, we construct a graph based on the similarity between tweets, and the graph can be divided into different event clusters to complete the event detection. Our method has achieved good results and can be applied to event detection in actual life.