Font Size: a A A

Research On Recognition Of Mental Diseases And Recommendation Of Mental Medical Resources Based On RoBERTa And Knowledge Graph

Posted on:2024-05-09Degree:MasterType:Thesis
Country:ChinaCandidate:Y XiaFull Text:PDF
GTID:2544307118953289Subject:Computer technology
Abstract/Summary:PDF Full Text Request
According to various data,the mental health problems of adolescents have become increasingly serious in recent years.The identification of traditional mental illness requires psychological counseling in a psychological hospital,and some patients have stigmatic psychology,patients and their relatives lack professional knowledge related to mental medicine,and mental illness itself is complex and hidden,and many reasons make mental illness difficult to be detected and intervened early.With the rapid development of artificial intelligence technology,artificial intelligence-related technologies are now used to solve the problem of mental illness recognition.In the field of mental illness recognition,this paper introduces the RoBERTa-TextGCN model to classify user conversations to identify users’ potential mental diseases,and then introduces knowledge graph to improve the recognition effect of mental illness and provide users with mental medical resource information.The main contributions of this study are as follows:(1)In this paper,a mental disease recognition model based on RoBERta-Text CNN is established,text classification models for mental illness and suicidal tendencies are established on the efaqa-corpus-zh psychological dataset,and comparative experiments are carried out with several basic models.The experimental results show that the RoBERTaText CNN model is superior to other models,and the accuracy of the model in the classification tasks of mental illness and suicidal tendency reaches 0.889 and 0.978,respectively.Experiments show the effectiveness and feasibility of the RoBERTa-Text CNN classificsation model in the field of mental disease recognition.(2)Aiming at the shortcomings of traditional neural networks such as Text CNN in extracting insufficient heterogeneous information and lacking global word co-occurrence information,this paper introduces a graph convolutional neural network,proposes a RoBERTa-TextGCN classification model,establishes a text heterogeneous graph on the same corpus dataset and conducts experiments,and the results show that RoBERTaTextGCN improves the accuracy of mental illness and suicidal tendency classification tasks by 1.38% compared with RoBERTa-Text CNN.0.73%.This shows that the RoBERTaTextGCN classification model has a certain improvement in training accuracy.(3)Using a top-down approach to construct a knowledge graph of mental medical resources,using crawler technology to obtain data from authoritative websites related to mental medicine,designing an ontology library,including mental illness,doctors,drugs,and hospital ontology,as well as ontology attributes and relationships,establishing a BERTBi LSTM-CRF entity recognition model,extracting entities in the mental medical field in the dataset,and completing the construction of a knowledge graph of mental health resources with the help of Neo4 j graph database.The operation mechanism of mental illness identification and psychological medical resource recommendation based on knowledge graph is analyzed.(4)On the basis of the previous research,from the perspective of practical engineering application,two application services are established,one is to build a mental illness recognition application service based on the RoBERTa-TextGCN model and the knowledge graph of psychological medical resources,first classify the user dialogue to initially identify the user’s potential mental illness,and then improve the mental illness recognition effect based on the operation mechanism of the knowledge graph.The second is to construct a mental illness medical resource recommendation service based on the knowledge graph of mental medical resources,and obtain recommendation information related to mental health resources based on the user information obtained by reasoning.In this paper,artificial intelligence technology is introduced into the field of mental health,mental illness identification model and mental health resource knowledge map are constructed,and certain work is done in the direction of mental illness identification and mental health resource recommendation of the adolescent mental health dialogue platform project,so as to provide help and research reference for solving the adolescent mental health problems.
Keywords/Search Tags:Identification of mental illness, Resource recommendation, Text classification, Knowledge graph, RoBERTa-TextGCN
PDF Full Text Request
Related items