Research On Convolution Event Extraction Of Syntactic Text Graphs Fused With Extended Words

In recent years,with the rapid development of resource demand in various fields of society,the development of information technology has become in full swing.The rapid development of information technology has led to a large amount of irregularly structured information displayed in various fields in the form of text.Excessive unstructured text information has no corresponding technology for processing and utilization,resulting in the loss of too much knowledge and value,so information extraction technology came into being.Information extraction technology aims to extract valuable information from unstructured and disorderly original data.Event extraction is a very important subtask in the task of information extraction.It can structure the information and knowledge in the text in a structured and comprehensive manner.The interpretation of the role appears.Event extraction technology extracts specific events and event instances from the text,and stores the complete structure of the event in the knowledge base.The main research content of this topic is the whole process of event clustering,which is mainly divided into two modules: trigger word recognition and argument classification.Specifically,it is the key steps involved in the process of event extraction and the process of linking to the knowledge graph.Mainly study the following contents:(1)Event trigger word detection classification and parameter recognition: The trigger word is an important role in identifying the type of event,and its positioning represents the event category and number of events contained in the entire sentence fragment.Trigger words are not only closely related to the recognition of event types,but also deeply affect the extraction of subsequent arguments and the construction of the entire event framework.This paper studies the operation of expanding trigger words and text graph convolution to improve the accuracy of identifying trigger words in the text and enhance the effect of event recognition and classification.(2)Research on event argument role classification extraction and structured extraction:After analyzing and identifying the trigger words and event types in the text,syntactic dependency analysis and improved kmeans clustering algorithm are used to extract the trigger words and trigger words in the text segment.Have connections and participate in an important role in the event instance.The composition and structural integrity of the event are not only related to the type of event that triggers word recognition,but also related to the degree of extraction of argument roles.Arguments related to trigger words are extracted a lot,which can improve the structure of the entire event instance,allowing humans to realize the visibility of the entire event instance more clearly and intuitively.The experimental results of this subject are divided into two modules,which are the experiment of extracting event trigger words through graph convolution and their recognition effect and the experiment of analyzing the syntactic dependence and improving the k-means algorithm to identify the argument composition of each event type..Experimental results show that in terms of event trigger word extraction,in the ACE corpus,the effect of event trigger word recognition through graph convolution is relatively good.In the CEC corpus,expanding the trigger vocabulary and then performing graph convolution semantic learning can gain more Good experimental results.In terms of argument recognition,the sentence structure is analyzed through syntactic dependence,and then the improved k-means algorithm is used to identify arguments.Compared with the effect of the k-means algorithm before the improvement,it is more accurate and achieves more economical equipment performance.Utilization rate,the identification effect of each element can be greatly improved.