Design And Implementation Of Automatic Question Answering System Based On Thyroid Knowledge Graph

With the improvement of people's living standards,the occurrence of thyroid disease has not been uncommon.However,the uneven distribution of thyroid medical resources in China has caused problems such as overcrowding in large hospitals,frequent patient attendance by doctors,and long patient visits.With the rapid development of “Internet Plus” technology and smart medical care,thyroid patients will seek online thyroid counseling platforms such as medical websites and online doctors' websites for thyroid disease consultation.However,this type of platform requires doctors to provide counseling services to patients through online manual consultations.This type of online consultation platform lacks automatic and intelligent ways of asking and answering patients,and it cannot provide prompt counseling for a large number of patients.How to provide an automated online question and answer service for thyroid patients has become a topic of widespread concern in the field of smart medicine.Thyroid patients will produce a large number of thyroid electronic medical records during the visit.These data provide a source of data for the implementation of an automated question and answer system for thyroid diagnosis and treatment.To this end,this paper builds a thyroid knowledge graph based on a thyroid electronic medical record from a top-three hospital in Shanghai.Based on thyroid knowledge graph,natural language processing and knowledge graph query techniques are used to design and implement an automated question answering system for thyroid diagnosis and treatment.The research content of this article mainly includes:1)Designed an overall framework of an automated question answering system based on thyroid knowledge graphs.The overall architecture of the system was then elaborated.The entire system was divided into the thyroid knowledge graph construction subsystem and the thyroid diagnosis and treatment automated question and answer subsystem thyroid knowledge graph construction subsystem responsible for the construction of thyroid diagnosis and treatment.The automated question answering system queries the knowledge base,and the thyroid diagnosis and treatment automated question and answer subsystem is responsible for translating the natural language question input by the user into a knowledge graph query statement and then obtaining the answer to the question sentence.Based on the various functions of the subsystems and the relationship between them,the architecture diagram of the entire system was designed,and the two subsystems were separately outlined.2)Secondly,the thyroid knowledge graph construction subsystem was introduced in detail: Firstly,the characteristics of thyroid electronic medical record data were analyzed,thyroid related terms were extracted,and related concepts of thyroid knowledge graph were obtained by synthesizing similar terms and the thyroid was designed accordingly.Knowledge graph conceptual model structure.Then,the concept mode structure is analyzed and the relationship between concepts is defined to complete the conceptual model design of the thyroid knowledge graph.Subsequently,thyroid-related data is extracted from the database as a set of entities,and entity filling operations are performed according to the conceptual model of the design.Finally,entity and entity relationships are formed in the form of triples <entity-relationship-attribute> to constitute a thyroid knowledge graph.3)An automated question-and-answer process flow based on thyroid knowledge graphs was designed: The thyroid diagnosis and treatment automated question-answering subsystem consists of a problem preprocessing module and an answer generation module.In the problem preprocessing module,the Chinese word segmentation algorithm is first used to segment the user-entered question,and the category of the question sentence is obtained through the keyword.Secondly,a LSTM+CRF algorithm was used to generate a recognition model for the thyroid consultation question corpus.This model was used to obtain thyroid entities from questions.Then,the LTP-parser tool was used to analyze the interjection syntax of the question sentence,and the subject-object relationship of each structure in the sentence was obtained.The trigram form of the question triple was formed.Finally,the entity of the question triple is mapped to the entity of the knowledge graph to avoid querying the entity that does not exist in the knowledge graph and obtain the query triple.In the answer generation module,the SPARQL query template is matched according to the question category,the problem of natural language is converted into the knowledge graph query language,and the query is answered through the knowledge graph to obtain the answer to the question sentence,and then the answer to the question sentence is targeted and eventually fed back to the user.4)An automated question answering system based on thyroid knowledge graphs is presented: First,the details of the construction of the thyroid knowledge graph and the construction results are displayed.Secondly,the implementation process of the question preprocessing module and the answer generation module in the automatic question answering system and the system achievements are displayed.The test results prove that the automatic question answering system in this paper has better usability.