| Summary: | With the rapid development of the Internet, social media is becoming more and more dominating in people’s daily life. On social media, users report and share their real-time observations with the world. Therefore, it can be used as a good source of real-time or close-to-real-time information. This information has great value and can be used in various ways. In this project, by using different artificial intelligence techniques, Twitter posts will be used to detect hazardous event. The whole project is divided into four parts. The first part is to extract data from twitter using twitter API for further use. The second part is classification. Using Long Short-Term Memory (LSTM) Model, posts are classified into different classes, such as earthquake, typhoon and shooting. Thirdly, after classification, unsupervised learning is performed to cluster posts related to a same event together. The posts in each class are embedded into Term Frequency–Inverse Document Frequency (TF- IDF) vectors and then clustered using K-Means Method. At last, for each cluster, we will use Spacy Named Entity Recognition method to extract useful information such as date, time and location to represent the hazardous event.
|
|---|