Research And Application Of Chinese Multi-relation Extraction Based On Fusion Model

In recent years,with the development of network technology and the popularity of network terminals,more and more scenes in life can be online.Take online education for example.The number of users has increased significantly in recent years.In particular,under the influence of COVID-19 in 2020,the online education business will grow rapidly,showing a good industry prospect.A good education platform is of great help to both students and teachers.However,the current online education platform still has some shortcomings,such as the high price of courses,the lack of recommendation ability or the recommendation content is not intelligent.In order to make up for the short board,the author tried to apply knowledge mapping,cognitive diagnosis technologies such as online education,this article preliminary build the knowledge map,education and other students cooperation to develop a deep and auxiliary AI(Artificial Intelligence)education system based on knowledge atlas,more accurate recommend conforms to the students to the purpose of learning content,in order to make more high quality online education platform.The main work of this thesis is as follows:1.In view of the problem of Chinese multi-relation extraction,BCMRE(Base-BERT Chinese Multi-relation extraction)model based on Bert(Bidirectional Encoder Representation from Transformers)is proposed,which can complete the extraction of multiple triples according to a single text sentence.This model is mainly composed of relational classification model and element extraction model.The relationship classification model is used to classify various relationships of sentences,and the element extraction model is used to identify named entities and get corresponding entities.In order to better classify the relational classification model,the premodel AGCNN(attention-gated convolutional neural network)was added,while in the element extraction model,the premodel Bi LSTM(Bi-directional Long Short-Term Memory)and CRF(Conditional Random Fields)were added for word segmentation and part of speech information classification.Finally,the validity of the model is verified by experiments.2.Clamping education data,using BCMRE to extract triples,and using Neo4 j visual database to create the knowledge graph and apply it to the project.Chinese knowledge triples were obtained from BCMRE,and then these triples were constructed into graph database through Neo4 j using Python script code written by the author,namely,knowledge graph.The design and development of deep assisted AI education system is completed based on knowledge graph.The innovation of the program is the integration of knowledge graph,cognitive diagnosis,recommendation and other AI technologies as the core of the program.Then,through demand analysis,development and testing,the online education auxiliary system is completed.