Netinfo Security ›› 2018, Vol. 18 ›› Issue (5): 75-81.doi: 10.3969/j.issn.1671-1122.2018.05.009

• Orginal Article • Previous Articles     Next Articles

Chinese Event Detection Based on Recurrent Neural Network

Chenxi MA1,2, Xingshu CHEN2,3, Wenxian WANG2,3, Haizhou WANG2,3()   

  1. 1. Network and Trusted Computing Institute, Computer College, Sichuan University, Chengdu Sichuan 610065, China
    2.Cybersecurity Research Institute, Sichuan University, Chengdu Sichuan 610065, China
    3.College of Cybersecurity, Sichuan University, Chengdu Sichuan 610065, China
  • Received:2017-11-26 Online:2018-05-15 Published:2020-05-11

Abstract:

With the development of Internet, the size of the Internet users has grown rapidly. The Internet has become more and more important to people’s life and social influence. In the face of the growing mass of Internet information, it is vital to quickly locate the events of public discussion. Event extraction is an important research in the field of information extraction. Event detection is the first step in the event extraction task, which plays a crucial role in the event extraction task.We designed a joint model based on recurrent neural network, to realize the recognition of event trigger and the classification of event category. Compared with the traditional method, our joint model can avoid error propagation, it doesn’t depend on the table of the trigger word and has good portability, and doesn’t need to design complex linguistic features.We used CEC corpus as training corpus and test corpus. The experimental results show that accuracy rate of the trigger word and event category is high, and the F value is 70.2%, better than the traditional method.

Key words: event detection, trigger, event extraction, RNN, word embedding

CLC Number: