Research On Entity Linking And Prediction Methods For Marine Economic Industry

In recent years,with the development of big data and artificial intelligence technology,it has become feasible to automatically acquire knowledge from large-scale multi-source heterogeneous data to build knowledge bases in various fields.Various industries and industries are actively promoting the construction of relevant knowledge graphs to support various applications of the industry in knowledge question and answer,fault diagnosis,and personalized recommendation.At present,the large-scale semantic knowledge base in marine economic industry is still blank,which needs to be studied urgently.Entity links are not accurate and complete,and the construction of marine knowledge base needs to be improved.Based on the existing algorithm,this paper strengthens the effect of entity linking compared with the previous entity linking framework,combined with the knowledge graph method.This article studies the entity link relationship prediction model and improves the entity link relationship prediction model.The main research contents are as follows:1)In view of the various types of marine data text entities,this paper uses a nested neural network for Chinese named entity recognition,and adds the Attention Mechanism to the previous named entity recognition framework,so that the model pays more attention to the integrity of the marine data text.The Chinese named entity recognition of marine data text is carried out by using BERT pre-training model and neural network,and the BIO mode is used to mark each entity,and then the place name,person name,organization name,etc.of the marine industry data are obtained,so as to achieve the effect of recognizing marine data entities.2)In order to solve the problem of ocean entity linking,this paper combines the knowledge graph technology and proposes an entity disambiguation framework to disambiguate ocean data.The frame work uses knowledge graph node embedding to vectorize entities,and uses word embedding methods to represent the entities of ocean data text.In the process of processing marine data text,this paper adds an improved ALBERT model,which selects the most suitable candidate entity set after similarity calculation,which can effectively solve the problem of marine economic industry entity disambiguation.3)In order to improve the knowledge base of the ocean knowledge graph,there will be the problem of missing the edges of the ocean knowledge graph.This paper uses two kinds of constraints to predict the entity link relationship,and uses the idea of single-step hierarchical constraints and multi-step hierarchical constraints to experiment with relational entities.In order to balance the difference between the positive and negative triples of the data text in the marine data industry,this paper improves the shortcomings of the hyperplane translation model,thereby perfecting the large-scale semantic construction of the marine knowledge graph.This paper proposes a set of entity linking process to improve and supplement the knowledge graph of marine industry.Experiments show that the nested neural network used in this paper is effective in Chinese named entity recognition.On the other hand,the combination of knowledge graph technology has a significant effect on entity disambiguation,making the entire disambiguation framework more complete.In the prediction of entity link relationship,the potential relationship between many entities is also improved,so that the entire ocean knowledge graph can be supplemented.This paper has promoted the construction of a large-scale semantic knowledge base for the marine economic industry,promoted the development of the marine economic industry,and achieved good results.