A medical consultation information recommendation method and system

Abstract

The application provides a medical consultation information recommendation method and system, which comprises the following steps: obtaining consultation information provided by a user, the consultation information at least comprising text data; determining a corresponding consultation question category according to the text data; inputting the text data into an automatic recognition model to obtain target data in the text data; obtaining a knowledge graph database; matching corresponding recommendation information in the knowledge graph database according to the target data and the consultation question category; and returning the recommendation information to the user to provide convenience for the user and improve the user's medical satisfaction.

Description

Technical Field

[0001] This invention belongs to the technical field of smart healthcare, specifically relating to a method and system for recommending medical consultation information. Background Technology

[0002] With the development of artificial intelligence (AI) technology, it has been integrated into various industries. Among them, AI technology is gradually being widely applied in the medical industry. It can be seen that smart healthcare based on the Internet is developing rapidly. Due to the special nature of the industry, compared with deep learning-based model technology, knowledge graph-based smart healthcare technology has interpretability and is being accepted and recognized by more and more industry professionals.

[0003] With the development of society, people's pace of life is getting faster and faster, and all walks of life are focusing on efficiency. In order to improve the efficiency of hospital visits, a reception desk is usually set up, where relevant staff provide services to patients. However, the number of relevant staff is limited, especially when there are many patients, it is impossible to attend to everyone. Moreover, the knowledge level of the relevant staff is limited, and they may not be able to provide the most appropriate answers to the questions raised by patients. Therefore, there is an urgent need for a method of recommending medical consultation information to facilitate patients' medical treatment and improve their satisfaction with medical treatment. Summary of the Invention

[0004] Based on this, the present invention provides a method and system for recommending medical consultation information, which aims to provide convenience for users seeking medical treatment and improve user satisfaction with medical treatment.

[0005] A first aspect of this invention provides a method for recommending medical consultation information, the method comprising:

[0006] Obtain consultation and diagnosis information provided by users, the consultation and diagnosis information including at least text data, and determine the corresponding consultation question category based on the text data;

[0007] The text data is input into the automatic recognition model to obtain the target data in the text data;

[0008] Obtain a knowledge graph database, match corresponding recommendation information in the knowledge graph database based on the target data and the category of the consultation question, and return the recommendation information to the user.

[0009] Furthermore, the step of inputting the text data into the automatic recognition model to obtain the target data in the text data includes:

[0010] Medical knowledge sample data is acquired and preprocessed to obtain target annotation data in the medical knowledge sample data, wherein the target annotation data includes at least entity data, attribute data corresponding to the entity data, and relation data;

[0011] Based on the entity data and the relation data, the medical knowledge sample data is trained to extract target data in order to establish a target data extraction sub-model.

[0012] The entity data and the relation data are combined to form the smallest unit of data storage, triplet data. Based on the smallest unit triplet data, relation extraction training is performed on the medical knowledge sample data to establish a relation extraction sub-model.

[0013] Based on the attribute data, attribute extraction training is performed on the medical knowledge sample data to establish an attribute extraction sub-model.

[0014] The target data extraction sub-model, the relationship extraction sub-model, and the attribute extraction sub-model constitute the automatic identification model.

[0015] Furthermore, before the step of acquiring the knowledge graph database, matching corresponding recommendation information in the knowledge graph database based on the target data and the category of the consultation question, and returning the recommendation information to the user, the following steps are included:

[0016] Acquire medical knowledge data, input the medical knowledge data into the automatic recognition model, and obtain the first entity data, first attribute data, first relation data, and first triplet data composed of the first entity data and the first relation data of the medical knowledge data;

[0017] The quality of the first triplet data is evaluated to obtain the quality evaluation result, and it is determined whether the evaluation result is qualified.

[0018] If so, the first entity data, the first attribute data, and the first relation data corresponding to the first triplet data are stored in the knowledge graph database.

[0019] Furthermore, the target data extraction sub-model and the attribute extraction sub-model employ the BILSTTM-CRF labeled data training algorithm, wherein the formula for the BILSTTM-CRF labeled data training algorithm is:

[0020] score(x, y) = ∑ i Emit(x i y i )+Trans(y i-1 yi );

[0021] Where x represents the input text data, y represents the corresponding output label, and i represents the number of characters in the text data. i Let y represent the i-th word in the text data. i y represents the label of the i-th word in the text data. i-1 Let represent the label of the (i-1)th word in the text data, and let score(x, y) represent the score of the label after inputting the text data. Emit(x) i y i Trans(y) represents the probability value of outputting the label of the i-th word in the input text data when the i-th word is in the input text data. i-1 y i ) represents the transfer score value of the label of the (i-1)th word in the text data after inputting the label of the i-th word into the transfer score matrix and transferring it to the label of the i-th word in the text data.

[0022] Furthermore, the step of combining the entity data and the relation data to form the smallest unit of data storage, triplet data, and then training the medical knowledge sample data on relation extraction based on the smallest unit triplet data to establish a relation extraction sub-model includes:

[0023] Obtain word data from the medical knowledge sample data, and generate corresponding child nodes based on the word data. The child nodes are used to store relationships and the positions of the corresponding child words.

[0024] Based on the word data, a corresponding dependency structure is generated. The dependency structure is used to record the first part of speech of the word data, the second part of speech of the parent node, and the first relationship between the first part of speech and the second part of speech.

[0025] The word data in the medical knowledge sample data is looped through, and a second relationship between the word data is extracted, wherein the second relationship includes at least verb-object relationship, post-modifier verb-object relationship, and subject-verb-complement relationship of prepositional phrase;

[0026] Based on the dependency structure, determine the target word data in the second relation that conforms to the dependency structure.

[0027] Furthermore, the step of storing the first entity data, the first attribute data, and the first relation data corresponding to the first triplet data into the knowledge graph database includes:

[0028] Entity disambiguation is performed on all the first entity data in the medical knowledge data to obtain the similarity value between the first entity data, and it is determined whether the similarity value is greater than a threshold.

[0029] If so, then the first entity data whose similarity values ​​meet the requirements will be subjected to knowledge fusion.

[0030] Furthermore, the step of acquiring the knowledge graph database, matching corresponding recommendation information in the knowledge graph database based on the target data and the category of the consultation question, and returning the recommendation information to the user includes:

[0031] Based on the recommended information, the corresponding department information and the information of the doctor currently on duty in the department information are matched, wherein the doctor information includes at least the doctor's name;

[0032] Based on the doctor's name, obtain the corresponding number of appointments, and push the doctor's name with the smallest number of appointments to the user.

[0033] A second aspect of this invention provides a recommendation system for medical consultation information, the system comprising:

[0034] The first acquisition module is used to acquire consultation and diagnosis information provided by the user, the consultation and diagnosis information including at least text data, and to determine the corresponding consultation question category based on the text data;

[0035] The target data acquisition module is used to input the text data into the automatic recognition model to obtain the target data in the text data;

[0036] The matching module is used to obtain a knowledge graph database, match corresponding recommendation information in the knowledge graph database according to the target data and the category of the consultation question, and return the recommendation information to the user.

[0037] A third aspect of the present invention provides a computer-readable storage medium, comprising:

[0038] The readable storage medium stores one or more programs that, when executed by a processor, implement a method for recommending medical consultation information in the first aspect.

[0039] A fourth aspect of the present invention provides an electronic device, characterized in that the electronic device includes a memory and a processor, wherein:

[0040] The memory is used to store computer programs;

[0041] When the processor executes the computer program stored in the memory, it implements the method for recommending medical consultation information in the first aspect.

[0042] This invention proposes a method and system for recommending medical consultation information. The method involves acquiring consultation and diagnosis information provided by a user, which includes at least text data. Based on the text data, the corresponding consultation question category is determined. The text data is then input into an automatic recognition model to obtain target data within the text data. A knowledge graph database is then accessed, and based on the target data and the consultation question category, corresponding recommendation information is matched within the knowledge graph database. This recommendation information is then returned to the user to facilitate medical treatment and improve user satisfaction. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the implementation of a method for recommending medical consultation information provided in the first embodiment of the present invention.

[0044] Figure 2 This is a flowchart illustrating the implementation of a medical consultation information recommendation system provided in the third embodiment of the present invention.

[0045] Figure 3 This is a structural block diagram of an electronic device provided in the fourth embodiment of the present invention.

[0046] The following detailed embodiments will be further described in conjunction with the above-mentioned accompanying drawings. Detailed Implementation

[0047] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0048] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0049] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0050] Example 1

[0051] Please see Figure 1, Figure 1 The present invention illustrates a method for recommending medical consultation information according to a first embodiment of the present invention, the method specifically including steps S01 to S03.

[0052] Step S01: Obtain consultation and diagnosis information provided by the user, the consultation and diagnosis information including at least text data, and determine the corresponding consultation question category based on the text data.

[0053] In this embodiment, a user terminal and a doctor terminal are provided. The user terminal is mainly for users to log in and register. The registration information includes the user's basic personal information. This information is stored in the information database during the registration process. The data is saved by the hospital's internal system to ensure the security of user privacy.

[0054] Specifically, the process begins by acquiring user-provided consultation information, which represents the questions users want to ask based on their needs. This consultation information can be text data (written statements) or image data (photos). In this embodiment, based on the user-provided text statements, a question classification algorithm model is used to categorize the consultation questions and determine their categories. The steps for establishing the question classification algorithm model are as follows: first, a dataset is collected, which includes at least various types of questions consulted by users; then, the questions are categorized as classification labels, for example, "medication questions," "surgical questions," etc. Further, the various types of questions in the dataset are then classified using a WORK algorithm. The d2vec algorithm model is trained to generate a vocabulary vector table. Undefined words can be represented using UNK. Furthermore, the dataset corresponding to the constructed questions and labels is preprocessed by segmenting sentences, removing stop words and low-frequency words, and converting words into vector representations. Labels are also digitally represented. During model training, parameters are set, including the number of convolutional layers, number of iterations, and learning rate. The dataset is divided into training and testing sets, and the data is vectorized and labeled before being used as input to the model for training and testing. The optimal model is selected and saved. Finally, the model is deployed as a service, calling the corresponding inference model to return the label category.

[0055] In addition, in some other optional embodiments, consultation and diagnosis information containing text, photos and voice can be sent directly to the doctor's end, that is, providing the user with a function window to describe their own symptoms, so that they can remotely obtain professional doctors' diagnosis, treatment and consultation. The doctor will diagnose the condition and provide feedback, and decide whether hospitalization is necessary, which can avoid the user's back and forth.

[0056] Step S02: Input the text data into the automatic recognition model to obtain the target data in the text data.

[0057] It should be noted that an automatic identification model needs to be established first. The steps for establishing the automatic identification model are as follows: acquiring medical knowledge sample data, namely, data from electronic materials in the disease field, medical books, medical website crawlers, etc., collecting and summarizing medical knowledge data sources as sample data, and preprocessing the medical knowledge sample data to obtain target labeled data in the medical knowledge sample data. The target labeled data includes at least entity data, attribute data corresponding to the entity data, and relation data. Specifically, the medical knowledge sample data can be labeled manually. The labeled objects can be mainly divided into entity classes, attribute classes, and relation classes. Entity classes can include diseases, drugs, treatment plans, surgical names, user symptoms, etc., while attribute classes can include indications, contraindications, usage and dosage, etc.

[0058] Furthermore, based on the labeled entity and relation data, target data extraction training is performed on the medical knowledge sample data to establish a target data extraction sub-model. Additionally, entity and relation data are combined to form the smallest unit of data storage: triplet data, such as <cold, fever, symptoms>. Here, "cold" is the defined disease entity, "fever" is the symptom entity, and "symptoms" represents the relationship between the two. Based on the smallest unit triplet data, relation extraction training is performed on the medical knowledge sample data to establish a relation extraction sub-model. More specifically, the steps for establishing this relation extraction sub-model are as follows: obtain medical knowledge samples... The system extracts word data from the sample data and generates corresponding child nodes based on the word data. These child nodes store relationships and the positions of the corresponding child words. It also generates corresponding dependency structures based on the word data, recording the first part of speech of the word data, the second part of speech of the parent node, and the first relationship between the first and second parts of speech. Finally, it extracts word data from the cyclical medical knowledge sample data and extracts the second relationships between the word data. These second relationships include at least verb-object relationships, post-modifier verb-object relationships, and subject-verb-verb-complement relationships. Based on the dependency structures, it identifies the target word data that conforms to the dependency structures within the second relationships.

[0059] Then, based on the attribute data, attribute extraction training is performed on the medical knowledge sample data to establish an attribute extraction sub-model. Understandably, the target data extraction sub-model, the relation extraction sub-model, and the attribute extraction sub-model constitute the automatic recognition model.

[0060] In this embodiment, the target data extraction sub-model, attribute extraction sub-model, and relation extraction sub-model all employ the Bilstm-Crf labeled data training algorithm. The formula for the Bilstm-Crf labeled data training algorithm is as follows:

[0061] score(x, y) = ∑ i Emit(xi， y i )+Trans(y i-1 y i );

[0062] Where x represents the input text data, y represents the corresponding output label, and i represents the number of characters in the text data. i Let y represent the i-th word in the text data. i y represents the label of the i-th word in the text data. i-1 Let represent the label of the (i-1)th word in the text data, and let score(x, y) represent the score of the label after inputting the text data. Emit(x) i y i Trans(y) represents the probability value of outputting the label of the i-th word in the input text data when the i-th word is in the input text data. i-1 y i ) represents the transfer score value of the label of the (i-1)th word in the text data after inputting the label of the i-th word into the transfer score matrix and transferring it to the label of the i-th word in the text data.

[0063] Step S03: Obtain the knowledge graph database, match corresponding recommendation information in the knowledge graph database according to the target data and the consultation question category, and return the recommendation information to the user.

[0064] Before acquiring the knowledge graph database, its establishment should be completed. Specifically, all available medical knowledge data should be acquired, such as medical books, expert experience, and electronic medical records. This medical knowledge data should be input into an automatic recognition model to obtain the first entity data, first attribute data, first relation data, and the first triplet data composed of the first entity data and the first relation data. Then, the quality of the first triplet data should be evaluated to obtain the quality evaluation result and determine whether the evaluation result is qualified. If so, the first entity data, first attribute data, and first relation data corresponding to the first triplet data should be stored in the knowledge graph database. It should be noted that the quality evaluation can be reviewed by experts in relevant fields. The final result data, i.e., the first triplet data, should be evaluated, and qualified data should be put into the knowledge graph database. The purpose is to ensure the quality of the knowledge graph database by discarding knowledge with low confidence.

[0065] Furthermore, knowledge typically exists in a dispersed, heterogeneous, and autonomous form, and also exhibits characteristics such as redundancy, noise, uncertainty, and incompleteness. Cleaning methods cannot solve these problems. Therefore, knowledge from related fields can be integrated, entity disambiguation and entity alignment can be performed, and finally input into a knowledge graph database. Entity disambiguation mainly calculates the similarity between entities and uses the entity similarity ranking to determine whether entities may overlap. In this embodiment, all first entity data in the medical knowledge data are subjected to entity disambiguation to obtain similarity values ​​between the first entity data, and it is determined whether the similarity value is greater than a threshold. If so, the first entity data with similarity values ​​that meet the requirements are then subjected to knowledge fusion.

[0066] Once the knowledge graph database is built, the system searches the database based on the target data and the type of inquiry. The results are then matched with the corresponding question template, and the language content is integrated to form a coherent description, which is then returned to the user.

[0067] Taking the inquiry "What are the symptoms of a cold?" as an example, the text data is input into the automatic recognition model. The target data extraction sub-model can identify the symptom entity as "cold". The relation extraction sub-model can determine that the inquiry is about "symptoms". Then, it matches the corresponding knowledge graph database query template, fills in the variable data, and converts it into a knowledge graph database statement that can be executed. Finally, it returns all the relevant symptoms of a cold, including "cough", "fever", and "runny nose". Then, through the question and answer recommendation information template, it is converted into fluent and understandable recommendation information, namely "The main symptoms of a cold are cough, fever and runny nose".

[0068] In summary, this invention obtains consultation and diagnosis information provided by users, which includes at least text data, and determines the corresponding consultation question category based on the text data; inputs the text data into an automatic recognition model to obtain target data from the text data; obtains a knowledge graph database, matches corresponding recommendation information in the knowledge graph database based on the target data and consultation question category, and returns the recommendation information to the user, thereby facilitating the user's medical treatment and improving the user's satisfaction with medical treatment.

[0069] Example 2

[0070] The second embodiment of the present invention provides a method for recommending medical consultation information, the method specifically including steps S10 to S12.

[0071] Step S10: Obtain consultation and diagnosis information provided by the user, the consultation and diagnosis information including at least text data, and determine the corresponding consultation question category based on the text data.

[0072] Step S11: Input the text data into the automatic recognition model to obtain the target data in the text data.

[0073] Step S12: Obtain the knowledge graph database, match the corresponding recommendation information in the knowledge graph database according to the target data and the consultation question category, and return the recommendation information to the user.

[0074] Specifically, based on the recommended information, the system matches the corresponding department information and the currently on-duty doctor information within that department. The doctor information includes at least the doctor's name and specialty. It's important to note that a mapping relationship between department information and recommended information should be established beforehand. For example, if the recommended symptom entity is "cold," then the user should consult an internal medicine doctor; thus, a mapping relationship is established between "cold" and internal medicine. If, in addition to "cold," symptoms such as "fever," "cough," "sputum," and "sore throat" are identified, a mapping relationship can be further established between "cold" and respiratory medicine. Similarly, if, in addition to "cold," symptoms such as "nausea," "vomiting," and "diarrhea" are identified, a mapping relationship can be further established between "cold" and gastroenterology. This allows users to more accurately choose the appropriate department and specialist doctor.

[0075] In addition, based on the doctor's name, the system retrieves the corresponding number of appointments and pushes the doctor's name with the fewest appointments to the user. Understandably, the more appointments a user has, the longer the waiting time for medical treatment will be. To more intelligently shorten the waiting time, the system recommends doctors with fewer appointments to the user, allowing the user to choose an appointment based on their own situation.

[0076] In this embodiment, the symptom description information sent by the user can be directly sent to the doctor. At the same time, the doctor can also see the user's basic information, which is the basic information filled in by the user when registering. Based on their own clinical experience and combined with the corresponding recommendation information matched in the knowledge graph database, the doctor can make a final judgment on the user's consultation and then reply to the user with the corresponding diagnosis results, thereby completing the entire response process of remote diagnosis and treatment.

[0077] Example 3

[0078] Please see Figure 2 , Figure 2 This is a structural block diagram of a medical consultation information recommendation system provided in an embodiment of the present invention. The medical consultation information recommendation system 300 includes: a first acquisition module 31, a target data acquisition module 32, and a matching module 33, wherein:

[0079] The first acquisition module 31 is used to acquire consultation and diagnosis information provided by the user, the consultation and diagnosis information including at least text data, and to determine the corresponding consultation question category based on the text data;

[0080] The target data acquisition module 32 is used to input the text data into the automatic recognition model to obtain the target data in the text data;

[0081] The matching module 33 is used to obtain a knowledge graph database, match corresponding recommendation information in the knowledge graph database according to the target data and the category of the consultation question, and return the recommendation information to the user.

[0082] Furthermore, in some optional embodiments of the present invention, the medical consultation information recommendation system 300 further includes:

[0083] A preprocessing unit is used to acquire medical knowledge sample data and preprocess the medical knowledge sample data to obtain target annotation data in the medical knowledge sample data, wherein the target annotation data includes at least entity data, attribute data corresponding to the entity data, and relation data;

[0084] The target data extraction sub-model building unit is used to train the medical knowledge sample data for target data extraction based on the entity data and the relation data, so as to build a target data extraction sub-model.

[0085] The relation extraction sub-model building unit is used to combine the entity data and the relation data to form the smallest unit triple data for data storage, and to perform relation extraction training on the medical knowledge sample data based on the smallest unit triple data to build a relation extraction sub-model.

[0086] The attribute extraction sub-model unit is used to perform attribute extraction training on the medical knowledge sample data based on the attribute data, so as to establish an attribute extraction sub-model.

[0087] The target data extraction sub-model, the relation extraction sub-model, and the attribute extraction sub-model constitute the automatic recognition model. All three sub-models—target data extraction, attribute extraction, and relation extraction—use the BILSTM-CRF labeled data training algorithm. The formula for the BILSTM-CRF labeled data training algorithm is:

[0088] score(x, y) = ∑ i Emit(x i y i )+Trans(y i-1 y i ;

[0089] Where x represents the input text data, y represents the corresponding output label, and i represents the number of characters in the text data. i Let y represent the i-th word in the text data. i y represents the label of the i-th word in the text data. i-1 Let represent the label of the (i-1)th word in the text data, and let score(x, y) represent the score of the label after inputting the text data. Emit(x) i y i Trans(y) represents the probability value of outputting the label of the i-th word in the input text data when the i-th word is in the input text data. i-1 y i ) represents the transfer score value of the label of the (i-1)th word in the text data after inputting the label of the i-th word into the transfer score matrix and transferring it to the label of the i-th word in the text data.

[0090] Furthermore, in some optional embodiments of the present invention, the medical consultation information recommendation system 300 further includes:

[0091] The first acquisition unit is used to acquire medical knowledge data, input the medical knowledge data into the automatic recognition model, and obtain the first entity data, first attribute data, first relation data, and first triplet data composed of the first entity data and the first relation data of the medical knowledge data.

[0092] The first judgment unit is used to perform quality assessment on the first triplet data, obtain the quality assessment result, and determine whether the assessment result is qualified.

[0093] The knowledge graph database construction unit is used to store the first entity data, the first attribute data, and the first relation data corresponding to the first triplet data into the knowledge graph database when the evaluation result is deemed qualified.

[0094] Furthermore, in some optional embodiments of the present invention, the relation extraction sub-model building unit includes:

[0095] The child node generates a sub-unit, which is used to obtain word data in the medical knowledge sample data and generate corresponding child nodes based on the word data. The child node is used to store the relationship and the position of the corresponding child word.

[0096] A dependency structure generation subunit is used to generate a corresponding dependency structure based on the word data. The dependency structure is used to record the first part of speech of the word data, the second part of speech of the parent node, and the first relationship between the first part of speech and the second part of speech.

[0097] Extraction and generation subunits are used to loop through the word data in the medical knowledge sample data and extract the second relationship between the word data, wherein the second relationship includes at least verb-object relationship, post-modifier verb-object relationship and prepositional subject-verb-verb-complement relationship;

[0098] The target word data determination subunit is used to determine the target word data in the second relation that conforms to the dependency structure, based on the dependency structure.

[0099] Furthermore, in some optional embodiments of the present invention, the knowledge graph database construction unit includes:

[0100] The first judgment subunit is used to perform entity disambiguation on all the first entity data in the medical knowledge data to obtain the similarity value between the first entity data, and to determine whether the similarity value is greater than a threshold.

[0101] The knowledge fusion subunit is used to perform knowledge fusion on the first entity data whose similarity value meets the requirements when the similarity value is determined to be greater than the threshold.

[0102] Furthermore, in some optional embodiments of the present invention, the medical consultation information recommendation system 300 further includes:

[0103] The matching module is used to match the corresponding department information and the information of the doctor currently on duty in the department information according to the recommendation information, wherein the doctor information includes at least the doctor's name and the doctor's specialty.

[0104] The push module is used to obtain the corresponding number of appointments based on the doctor's name, and push the doctor's name with the smallest number of appointments to the user.

[0105] Example 4

[0106] In another aspect, the present invention also proposes an electronic device, please refer to [link to relevant documentation]. Figure 3 The diagram shown is a structural block diagram of an electronic device according to the fourth embodiment of the present invention, including a memory 20, a processor 10, and a computer program 30 stored in the memory and executable on the processor. When the processor 10 executes the computer program 30, it implements the method for recommending medical consultation information as described above.

[0107] In some embodiments, the processor 10 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 20 or process data, such as executing access restriction programs.

[0108] The memory 20 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 20 can be an internal storage unit of an electronic device, such as the hard disk of the electronic device. In other embodiments, the memory 20 can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. Furthermore, the memory 20 can include both internal and external storage units of the electronic device. The memory 20 can be used not only to store application software and various types of data of the electronic device, but also to temporarily store data that has been output or will be output.

[0109] It should be pointed out that, Figure 3 The structure shown does not constitute a limitation on the electronic device. In other embodiments, the electronic device may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0110] This invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method for recommending medical consultation information as described above.

[0111] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0112] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0113] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions for data states, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0114] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0115] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for recommending medical consultation information, characterized in that, The method includes: Obtain consultation and diagnosis information provided by users, the consultation and diagnosis information including at least text data, and determine the corresponding consultation question category based on the text data; The text data is input into the automatic recognition model to obtain the target data in the text data; Obtain a knowledge graph database, match corresponding recommendation information in the knowledge graph database based on the target data and the category of the consultation question, and return the recommendation information to the user; Before the step of inputting the text data into the automatic recognition model to obtain the target data in the text data, the following steps are included: Medical knowledge sample data is acquired and preprocessed to obtain target annotation data in the medical knowledge sample data, wherein the target annotation data includes at least entity data, attribute data corresponding to the entity data, and relation data; Based on the entity data and the relation data, the medical knowledge sample data is trained to extract target data in order to establish a target data extraction sub-model. The entity data and the relation data are combined to form the smallest unit of data storage, triplet data. Based on the smallest unit triplet data, relation extraction training is performed on the medical knowledge sample data to establish a relation extraction sub-model. Based on the attribute data, attribute extraction training is performed on the medical knowledge sample data to establish an attribute extraction sub-model. The target data extraction sub-model, the relationship extraction sub-model, and the attribute extraction sub-model constitute the automatic identification model. Before the step of acquiring the knowledge graph database, matching corresponding recommendation information in the knowledge graph database based on the target data and the category of the consultation question, and returning the recommendation information to the user, the following steps are included: Acquire medical knowledge data, input the medical knowledge data into the automatic recognition model, and obtain the first entity data, first attribute data, first relation data, and first triplet data composed of the first entity data and the first relation data of the medical knowledge data; The quality of the first triplet data is evaluated to obtain the quality evaluation result, and it is determined whether the evaluation result is qualified. If so, the first entity data, the first attribute data, and the first relation data corresponding to the first triplet data are stored in the knowledge graph database.

2. The method for recommending medical consultation information according to claim 1, characterized in that, The target data extraction sub-model, the attribute extraction sub-model, and the relation extraction sub-model all employ the BILSTTM-CRF labeled data training algorithm. The formula for the BILSTTM-CRF labeled data training algorithm is as follows: ； Where x represents the input text data, y represents the corresponding output label, and i represents the number of characters in the text data. i Let y represent the i-th word in the text data. i y represents the label of the i-th word in the text data. i-1 Let represent the label of the (i-1)th word in the text data, and let score(x, y) represent the score of the label after inputting the text data. Emit(x...) i y i Trans(y) represents the probability value of outputting the label of the i-th word in the input text data when the i-th word is in the input text data. i-1 y i ) represents the transfer score value of the label of the (i-1)th word in the text data after inputting the label of the i-th word into the transfer score matrix and transferring it to the label of the i-th word in the text data.

3. The method for recommending medical consultation information according to claim 2, characterized in that, The steps of combining the entity data and the relation data to form the smallest unit of data storage, triplet data, and training the medical knowledge sample data on relation extraction based on the smallest unit triplet data to establish a relation extraction sub-model include: Obtain word data from the medical knowledge sample data, and generate corresponding child nodes based on the word data. The child nodes are used to store relationships and the positions of the corresponding child words. Based on the word data, a corresponding dependency structure is generated. The dependency structure is used to record the first part of speech of the word data, the second part of speech of the parent node, and the first relationship between the first part of speech and the second part of speech. The word data in the medical knowledge sample data is looped through, and a second relationship between the word data is extracted, wherein the second relationship includes at least verb-object relationship, post-modifier verb-object relationship, and subject-verb-complement relationship of prepositional phrase; Based on the dependency structure, determine the target word data in the second relation that conforms to the dependency structure.

4. The method for recommending medical consultation information according to claim 3, characterized in that, The step of storing the first entity data, the first attribute data, and the first relation data corresponding to the first triplet data into the knowledge graph database includes: Entity disambiguation is performed on all the first entity data in the medical knowledge data to obtain the similarity value between the first entity data, and it is determined whether the similarity value is greater than a threshold. If so, then the first entity data whose similarity values ​​meet the requirements will be subjected to knowledge fusion.

5. The method for recommending medical consultation information according to claim 4, characterized in that, The step of acquiring a knowledge graph database, matching corresponding recommendation information in the knowledge graph database based on the target data and the category of the consultation question, and returning the recommendation information to the user includes the following: Based on the recommended information, the corresponding department information and the information of the doctor currently on duty in the department information are matched, wherein the doctor information includes at least the doctor's name; Based on the doctor's name, obtain the corresponding number of appointments, and push the doctor's name with the smallest number of appointments to the user.

6. A recommendation system for medical consultation information, characterized in that, The system is used to implement the method for recommending medical consultation information according to any one of claims 1-5, the system comprising: The first acquisition module is used to acquire consultation and diagnosis information provided by the user, the consultation and diagnosis information including at least text data, and to determine the corresponding consultation question category based on the text data; The target data acquisition module is used to input the text data into the automatic recognition model to obtain the target data in the text data; The matching module is used to obtain a knowledge graph database, match corresponding recommendation information in the knowledge graph database according to the target data and the category of the consultation question, and return the recommendation information to the user.

7. A computer-readable storage medium, characterized in that, include: The readable storage medium stores one or more programs that, when executed by a processor, implement the method for recommending medical consultation information as described in any one of claims 1-5.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the method for recommending medical consultation information as described in any one of claims 1-5.

Images