| Event detection is an important task in text information extraction.It is widely used in the real world and is also one of the most important research topics in natural language processing.In recent years,with the development of machine learning and deep learning technology,new progress has been made in event detection research;however,existing event detection methods ignore the use of event category information in the process of learning features.Based on this,this paper proposes two new event detection methods,which com Bine the consideration of event category information to improve the effect of event detection.(1)An event detection model based on attention mechanism fusion of category information is proposed.The model includes three modules: context information encoding module,category information encoding module,information fusion and prediction module.The input text is passed through the context information encoding module,and the word vector sequence based on the text context information is output and obtained.In the category information encoding module,the word vector layer converts the category set into a sequence of word vectors,that is,the vector representation of event category information.Then,the vector representation of textual context information is fused with the vector representation of event category information,and event detection is performed.The experimental results show that the event category information is helpful for event detection,and the effectiveness of the model is effectively improved after introducing the attention mechanism.(2)An event detection model based on graph neural network fusion of category-level information is proposed.The model adds a graph neural network to the previous model to model category-level information.In the module of extracting category information,a graph neural network layer is constructed based on the event category structure diagram.Then,the category level information is extracted based on graph neural network,and the vector representation corresponding to the event category level information is obtained.Finally,build the model and detect events.The experimental results show that the model plays an important role in the modeling of category structure information. |
