Construction And Analysis Of Expert Knowledge Graphs For The Needs Of Chemical Enterprises

In recent years,colleges and universities have been gathering places for experts to meet the needs of chemical companies for professional scientific research and innovative talents.However,how to make expert searched intelligent and intelligently feedback the results required by users according to users’ intentions need to be solved urgently.Assisted search technology based on knowledge graph has received wide attention in various industries,and applied research in computer,medical,agriculture and other fields.The core problem of the construction of knowledge graph lies in knowledge extraction.In dealing with unstructured data,the keyword-related knowledge system is systematized so that users can find more accurate information.In addition to the open domain-oriented knowledge graph,knowledge graph technology can be used to automatically extract and construct a specific field,so that the computer can have a deep understanding of the field and more accurately extract keyword-related knowledge in the field.This paper studies the above core issues,and builds an expert knowledge map for the needs of chemical companies in the chemical industry.The main research issues include:(1)Chemical unstructured text classification.The WSD hierarchical memory networks are proposed for the hierarchical structure of expert text.The classification method is to classify expert texts from the expressiveness and effectiveness of the model.(2)Extract chemical expert entity information.This paper proposes an expert entity extraction method in chemical industry,which based on multi-feature bidirectional neural network.This method effectively solves the problem that traditional expert named entity recognition methods rely too much on artificial feature annotation and word segmentation effect,and effectively identifies professional new words.(3)For the entity extraction ambiguity problem,on the one hand,we propose a hybrid feature fusion chemical text segmentation and relationship extraction method to accurately segment chemical keyword entities.On the other hand,we extract the relationship between entities based on statistical methods.This method solves the diversity problem of natural language description,effectively disambiguates entities,and improves the accuracy of knowledge graphs.