Design And Implementation Of Knowledge Extraction Algorithm Based On Natural Language Processing

Knowledge extraction refers to the technology of detecting,screening and analyzing data,acquiring the knowledge and storing it in the knowledge base.The intelligence analysis system needs to obtain the events and entities involved in the intelligence according to the intelligence category.At this time,knowledge extraction technology based on natural language processing can efficiently solve such problems.In order to realize the extraction and display of entities and events in the English intelligence analysis system,a knowledge extraction algorithm has been designed and implemented,and a complete application system was constructed by combining front and back ends and databases.First introduces the development background of the knowledge extraction algorithm,then combines the domestic and international overview of the knowledge extraction technology,summarizes the key technologies involved in the project implementation,and after the needs analysis and architecture design,designs and implements each knowledge extraction algorithm modules and build them into a service.The main work includes: analyzing project requirements,designing an intelligence information knowledge structure,designing and implementing a data preprocessing module to extract the structured intelligence information;designing and implementing the topic categorization module to get the text data of Navy from the cleaned data;designing and implementing a named entity recognition module based on bidirectional long-short-term memory network and conditional random fields to extract the name of the ship,the time of the incident,and the location of the incident from the intelligence information;designing and implementing event extraction module based on dynamic multi-pooling convolutional neural network,which combined with named entity recognition results to recognize and classify event trigger words and arguments,extract events that occur in intelligence information;integrating knowledge extraction algorithms into marine defense systems,building portable images,and deploying knowledge extraction services to remote servers.The English intelligence analysis system project has undergone detailed functional tests and performance tests.Among them,the F1 value of the topic categorization module was 90.6,the F1 value of the named entity recognition module reached 94.2,and the F1 value of the event extraction module was 82.9.The project has been used online by the demand unit,and it can automatically crawl data from the website for knowledge extraction and display it on the front-end webpage according to user operations.