| BackgroundWith the rapid development of information technology,the field of information services is changing towards integration,personalization,knowledge,and intelligence.At present,in the field of medical and health care,the way patients obtain health information services is still mainly in the form of patients taking the initiative to consult doctors to obtain health information.Although the Internet provides a huge amount of health information resources,for patients who are information disadvantaged,their limited level of health literacy prevents them from accurately expressing health information needs that can be identified by the information service system,and the huge amount of health information adds to their cognitive burden.Artificial intelligence technology has injected new energy into the information services sector.In the context of"Internet+Healthcare",the existing patient-oriented information service model needs to be transformed into a personalized and intelligent one.It is urgent to build an intelligent information service model that is data-driven,combined with artificial intelligence technology to accurately identify patients health information needs and organize multi-source health information resources at a finer granularity.ObjectiveTo build a patient-oriented intelligent information service model,taking patients health information needs as the starting and ending point,intelligently mining patients demand characteristics,organizing health knowledge at a finer granularity and dynamically,and developing the intelligent information service model into an intelligent information service system to achieve the effect of providing intelligent services of automatic question and answer and active recommendation for patients.MethodsFirstly,literature research was used to analyze the current situation of research on patients health information needs and intelligent informa tion services at home and abroad,and a theoretical framework of intelligent information services was constructed at the theoretical level.In the construction of the intelligent information service model,various methods were used to solve the key problems of patient health information demand feature mining and health knowledge representation.Firstly,the division of multi-level and multi-dimensional features of patient health information needs is analyzed using the theoretical analysis method;the multidimensional patient health information needs feature system is designed using the content analysis method;the RoBERTa+BiLSTM+CRF/Attention method is used to train the model on the needs feature annotation data and construct a multi-dimensional needs feature identification model,on which the multi-level Then,the knowledge element is introduced to design the health knowledge meta description model to describe the knowledge from medical encyclopedia entries,online Q&A case knowledge and clinical guidelines at a fine-grained level,and the dictionary matching and rule matching methods are used to realize the automatic extraction of health knowledge meta names,knowledge items,attributes and relationship types by the computer system,and the empirical research On this basis,theoretical analysis is used to analyze the problem of patients access to health information services and to propose an intelligent information service model for patients health information needs.Finally,the system engineering method was used to implement the intelligent information service model with hypertensive patients as the service target.ResultsThis study proposes a theoretical framework for intelligent information service,which means that the patients current health problem to be solved is expressed in natural language as a health information need,and the computer system can understand the natural language,accurately identify the problem and provide relevant knowledge services to the patient to meet the patients health information needs.Based on the research on automatic mining of patient health information needs,a four-level patient health information needs feature consisting of explicit,expressive implicit,cognitive implicit,and objective implicit is proposed,and multidimensional health information needs feature consisting of "context+question" that can be represented by text data at each level is proposed.A combined RoBERTa+BiLSTM+CRF/Attention model with optimal performance in identifying multidimensional demand features in an experimental environment is constructed,which can effectively identify multidimensional patient health information demand features automatically from the text of patient questions.The model RoBERTa+BiLSTM+CRF has the best performance in the background dimensional demand feature recognition task with an F1 value of 82.17%,and the model RoBERTa+BiLSTM+Attention has the best prediction performance in the question classification task with an F1 value of 86.28%.On this basis,a computer system was designed to carry out automatic mining ideas for multi-level patient health information needs features.A health knowledge meta description model was proposed based on knowledge element.The model contains 10 types of knowledge elements and 11 semantic relationships,and can aggregate encyclopedic knowledge,web-based Q&A case knowledge,and clinical guideline knowledge to form a multisource health knowledge network.The study proposes an intelligent information service model for patients health information needs.The model illustrates the relationships between the subject,object,technology and environment involved,and provides automatic Q&A and active recommendation information services to meet patients multi-level health information needs based on demand-knowledge matching calculations by mining the characteristics of patients multi-level health information needs and fine-grained representation of multi-source health knowledge.Finally,an intelligent information service system for hypertensive patients is implemented.The system collects the knowledge of hypertension-related web quiz cases,encyclopedic knowledge and hypertension prevention,and treatment guidelines.The system uses 8338 hypertension patients question text data from the Q&A case knowledge to construct a Word2vec word vector model that can calculate the semantic distance of demand feature label words,and an LDA topic model with a topic count of 221 based on the demand feature label set and synonym expansion.According to the topic distribution of different patients,it can be found that patients health information needs are personalized.A multi-source hypertension health knowledge network containing 16737 knowledge meta-nodes and 131,563 edges was constructed by conducting health knowledge meta-based hypertension-related knowledge aggregation.In the experimental environment,the system can provide automatic question-and-answer services and active recommendation knowledge services at different levels for hypertension patients.ConclusionThe intelligent information service model proposed in this study can effectively solve the problem of mining the demand features of the question text data described by patients in natural language,and can effectively solve the problem of constructing multi-source health knowledge networks and the problem of matching between multi-level patient demand and multi-source health knowledge.The intelligent information service system developed in this study can visually present patients’ multi-dimensional and multi-level health information demand characteristics and dynamically match their demand characteristics with health knowledge,providing automatic Q&A and active recommendation services.Innovation(1)This study innovatively used natural language understanding technology to design a multi-level patient health information demand feature mining idea.The multi-dimensional demand feature recognition model proposed in this study can achieve the goal of automatically extracting the patients health information demand features from the patients question text.Based on the text data of patients questions,this study innovatively proposes an automatic mining process of multi-level patient health information demand characteristics.The characteristics,the characteristics of patients health information needs in the cognitive layer,and the characteristics of patients health information in the objective layer are represented,which expands the research field of patients health information needs.(2)This study innovatively introduced the knowledge metatheoretical model to build a multi-source health knowledge network of medical encyclopedia entry knowledge,patient-doctor question and answer pair case knowledge,and medical guideline knowledge,and used automatic question answering and active recommendation services for hypertensive patients.The function is taken as an example to verify the effectiveness of the multi-source health knowledge representation.This study extends the application of knowledge metatheory in patient health information service scenarios.(3)This study designed an intelligent information service model for hypertensive patients based on real hypertensive patients question text data and implemented it systematically.The system can automatically mine and visualize the health information demand characteristics of patients at the explicit level,the characteristics of patients health information demand at the expression level,the characteristics of patients health information demand at the cognitive level,and the characteristics of patient health information at the objective level from the questioning texts of patients with hypertension.Actively recommending service methods to dynamically match the needs of hypertensive patients and health knowledge,to meet the health information needs of hypertensive patients,and provide a new idea for the study of patient-oriented intelligent information service systems.Figures:43,Tables:16,References:179... |
PDF Full Text Request