Intelligent guiding method and device

Abstract

One or more embodiments of the present application provide an intelligent guidance method and device, the method comprising: acquiring a query text for guidance consultation, and performing named entity recognition on the query text to identify named entities and their types from the query text; if the identified named entities include disease characteristics, inputting the disease characteristics into a trained prediction model, predicting the corresponding disease based on the disease characteristics by the prediction model, and determining the department corresponding to the disease based on a graph; the nodes in the graph include nodes representing consultation characteristics and nodes representing departments; the consultation characteristics represented by the nodes connected by each edge in the graph correspond to the departments represented by the nodes connected by the edge; the consultation characteristics at least include diseases; if the identified named entities include consultation characteristics, determine the department corresponding to the consultation characteristics based on the graph; input the department into a guidance large model, and generate an answer text based on the department by reasoning of the guidance large model.

Description

TECHNICAL FIELD

[0001] One or more embodiments of the present application relate to the technical field of artificial intelligence, and in particular, to an intelligent guiding method and device. BACKGROUND

[0002] In the current medical scene, an entrance of each medical institution is generally provided with a guiding desk to assist patients who feel confused due to lack of professional medical knowledge. These patients usually have doubts about which department they should choose to visit, and therefore need to consult at the guiding desk to obtain a department corresponding to their symptoms, diseases, inspection / examination items, etc.

[0003] From the perspective of patients, since the service personnel at the guiding desk are usually not enough, patients first need to queue up at the guiding desk to consult. Subsequently, patients also need to go through a series of processes such as registration, waiting, visiting, obtaining a prescription issued by a doctor, and paying fees. The whole process of visiting a doctor is complicated and often needs to be queued. This not only wastes a lot of time of patients, but also increases anxiety and inconvenience of patients. From the perspective of medical institutions, their human resources are limited, and it is easy to have a shortage of human resources.

[0004] In view of the above, it is generally expected that a part of manual services, especially guiding services which are not strongly related to real disease diagnosis, can be replaced by intelligent services. In this way, the medical process of patients can be simplified and efficiency can be improved, and medical institutions can also help to alleviate the pressure of human resources and optimize the quality of services, so as to help to improve the overall medical service experience. Therefore, in practical applications, how to provide intelligent guiding services for users has become a problem to be solved. SUMMARY

[0005] One or more embodiments of the present application provide technical solutions as follows:

[0006] The present application provides an intelligent guiding method, which comprises:

[0007] obtaining a query text for guiding consultation, and performing named entity recognition on the query text to identify named entities from the query text and determine types of the identified named entities;

[0008] If the identified named entity includes a disease feature named entity, the disease feature named entity is input into a trained prediction model, a corresponding disease named entity is predicted by the prediction model based on the disease feature named entity, and a department named entity corresponding to the disease named entity is determined based on a graph; wherein nodes in the graph include a treatment feature node and a department node, each treatment feature node represents a treatment feature named entity, and each department node represents a department named entity; edges in the graph are used to connect the treatment feature nodes and the department nodes, and a treatment feature named entity represented by a treatment feature node connected by each edge corresponds to a department named entity represented by a department node connected by the edge; the treatment feature named entity includes at least one sub-type of named entity, and the at least one sub-type includes a disease.

[0009] If the identified named entity includes the treatment feature named entity, a department named entity corresponding to the treatment feature named entity is determined based on the graph;

[0010] The department named entity is input into a large-scale guidance model, and the large-scale guidance model performs reasoning based on the department named entity to generate an answer text corresponding to the query text.

[0011] The present application also provides an intelligent guidance device, which comprises:

[0012] An acquisition module acquires a query text for guidance consultation, and performs named entity recognition on the query text to identify named entities from the query text and determine the types of the identified named entities;

[0013] A first determination module, if the identified named entity includes a disease feature named entity, the disease feature named entity is input into a trained prediction model, a corresponding disease named entity is predicted by the prediction model based on the disease feature named entity, and a department named entity corresponding to the disease named entity is determined based on a graph; wherein nodes in the graph include a treatment feature node and a department node, each treatment feature node represents a treatment feature named entity, and each department node represents a department named entity; edges in the graph are used to connect the treatment feature nodes and the department nodes, and a treatment feature named entity represented by a treatment feature node connected by each edge corresponds to a department named entity represented by a department node connected by the edge; the treatment feature named entity includes at least one sub-type of named entity, and the at least one sub-type includes a disease.

[0014] A second determination module, if the identified named entity includes the treatment feature named entity, a department named entity corresponding to the treatment feature named entity is determined based on the graph;

[0015] The reasoning module inputs the named entity of the department into the patient guidance model, and the patient guidance model performs reasoning based on the named entity of the department to generate the answer text corresponding to the query text.

[0016] This application also provides an electronic device, including:

[0017] processor;

[0018] Memory used to store processor-executable instructions;

[0019] The processor executes the executable instructions to implement the steps of the method as described in any of the preceding descriptions.

[0020] This application also provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method as described in any of the preceding claims.

[0021] In the above technical solution, a graph for patient guidance can be pre-constructed. The nodes in this graph include nodes representing named entities of patient visit features and nodes representing named entities of departments. The named entities of patient visit features represented by the nodes connected by the edges correspond to the named entities of departments represented by the nodes connected by those edges. Subsequently, when a query text for patient guidance consultation is obtained, named entity recognition can be performed on the query text to identify the named entities and their types. If the identified named entities include disease feature named entities, the disease feature named entities can be first input into a trained prediction model. The prediction model then predicts the corresponding disease named entities based on the disease feature named entities. Based on the graph, the department named entities corresponding to the disease named entities are then determined. If the identified named entities include patient visit feature named entities, the department named entities corresponding to the patient visit feature named entities can be directly determined based on the graph. These department named entities can then be input into a large-scale patient guidance model, which infers based on these department named entities to generate an answer text corresponding to the query text. This answer text can then be used for patient guidance.

[0022] By employing the above methods, intelligent triage services are realized. This simplifies the patient's medical process, improves efficiency, and helps medical institutions alleviate human resource pressure and optimize service quality, thereby contributing to an improved overall healthcare experience. Furthermore, in providing intelligent triage services, the trained prediction model and the constructed graph can be used to determine the corresponding departmental named entities based on various medical feature named entities in the query text used for triage consultation. These determined departmental named entities are then integrated into the actual input of the large-scale triage model. This allows the large-scale model to consider both its inherent generalization knowledge and the specific knowledge reflected by the departmental named entities when generating the answer text corresponding to the query text, thereby improving the adaptability and response accuracy of the large-scale triage model. Attached Figure Description

[0023] The accompanying drawings used in the description of the exemplary embodiments will now be explained, wherein:

[0024] Figure 1 This is a schematic diagram illustrating an intelligent dialogue system according to an exemplary embodiment of this application.

[0025] Figure 2 This is a schematic diagram illustrating an intelligent triage process according to an exemplary embodiment of this application.

[0026] Figure 3A This is a schematic diagram illustrating an exemplary embodiment of this application.

[0027] Figure 3B This is a schematic diagram illustrating another figure as shown in an exemplary embodiment of this application.

[0028] Figure 4 This is a flowchart illustrating an exemplary embodiment of the present application of an intelligent triage method.

[0029] Figure 5 This is a schematic diagram of the structure of a device shown in an exemplary embodiment of this application.

[0030] Figure 6 This is a block diagram illustrating an exemplary embodiment of the present application of a smart triage device. Detailed Implementation

[0031] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of this application. Rather, they are merely examples consistent with some aspects of one or more embodiments of this application.

[0032] It should be noted that the steps of the corresponding methods are not necessarily performed in the order shown and described in this application in other embodiments. In some other embodiments, the methods may include more or fewer steps than those described in this application. Furthermore, a single step described in this application may be broken down into multiple steps in other embodiments; and multiple steps described in this application may be combined into a single step in other embodiments.

[0033] In the field of medical consultation, intelligent dialogue systems have become a key tool for improving efficiency and service quality. These systems provide consultation services to patients by simulating human communication, including pre-visit guidance and during the consultation process, providing disease diagnosis, treatment suggestions, and medication instructions.

[0034] An intelligent dialogue system is an interactive system developed using artificial intelligence technology. It aims to understand and answer questions posed by users in natural language, generating concise and clear responses. Intelligent dialogue systems are typically based on large language models, which understand and respond to user questions, generating corresponding answers.

[0035] Large language models are deep learning models trained on large amounts of text data. They can be used to generate natural language text or understand the meaning of natural language text. Large language models can handle a variety of natural language tasks, such as text classification, named entity recognition (NER), and dialogue, and are an important pathway to artificial intelligence.

[0036] In the field of natural language processing, large-scale text datasets are often referred to as corpora. Corpora can contain various types of text data, such as literary works, academic papers, legal documents, news reports, everyday conversations, emails, and online forum posts. By learning from the text data in corpora, large language models can acquire and understand the rules and patterns of natural language, thereby achieving effective processing and generation of human language.

[0037] Large language models typically employ the Transformer architecture; that is, large language models are usually deep learning models based on the Transformer architecture. Deep learning models based on the Transformer architecture are a class of neural network models that utilize the Transformer architecture, and these models perform exceptionally well in fields such as natural language processing.

[0038] A Transformer is a neural network model for sequence-to-sequence modeling. The Transformer does not rely on a recurrent structure, and can parallelize training and inference, accelerating the model processing speed. In a deep learning model based on the Transformer architecture, a multi-layer Transformer encoder is usually used to extract features from an input sequence, and a Transformer decoder is used to convert the extracted features into an output sequence. At the same time, such a model usually also uses a self-attention mechanism to capture long-distance dependencies in the input sequence, and uses residual connections and normalization methods to speed up training and improve model performance.

[0039] A pre-trained model is a large language model pre-trained on large-scale unlabeled text data. The pre-trained model is a general-purpose model that is not designed and optimized for a specific task. In order to adapt the pre-trained model to specific application scenarios and task requirements, fine-tuning is needed to improve the model's performance on specific tasks. The final large language model used in practice is usually a model that is further fine-tuned based on a pre-trained model, and is a supervised learning model based on labeled text data. Pre-training and fine-tuning are complementary processes, pre-training enables the model to have broad language understanding capabilities, and fine-tuning makes the model more specialized and accurate on specific tasks.

[0040] That is, the training process of a large language model can be divided into two stages: pre-training and fine-tuning. In the pre-training stage, unsupervised learning (e.g., self-supervised learning) can be used to pre-train on large-scale, unlabeled text datasets (e.g., web encyclopedias, web articles, books, etc.). Specifically, the model can predict missing parts or the next word based on context, learn statistical rules and language structures such as semantics and syntax, and iteratively update model parameters by minimizing prediction loss through backpropagation and optimization algorithms (e.g., gradient descent) to gradually improve the model's understanding of language. In the fine-tuning stage, according to the specific application scenario and task requirements, a corresponding supervised learning task (e.g., text classification, named entity recognition, dialogue system, etc.) can be selected and a task-specific text dataset can be prepared. Thus, the pre-trained model can be used as the starting point for fine-tuning, and supervised learning can be used to fine-tune the model on the task-specific text dataset. Specifically, the model can perform the task based on the text dataset, and iteratively update the model parameters by minimizing the loss that measures the model's performance in handling specific tasks through backpropagation and optimization algorithms (e.g., gradient descent) to gradually improve the model's performance on specific tasks. In practical applications, fine-tuning can flexibly choose supervised learning, unsupervised learning, or semi-supervised learning based on the specific application scenario and the type of available data.

[0041] It should be noted that the pre-trained large language model is usually referred to as the base model of the large language model, and the fine-tuned large language model is referred to as the service model of the large language. The language understanding ability learned by the large language model in the pre-training stage and the fine-tuning stage enables the large language model to understand, analyze, and synthesize text information, and to have the ability to logically deduce, knowledge reason, or solve problems when facing complex problems or tasks. This ability is usually referred to as the reasoning ability of the large language model.

[0042] The large language model usually performs specific tasks under the guidance of a prompt text (which can be referred to as a Prompt). The prompt text is an initial text or text segment provided to the large language model to stimulate the model to generate corresponding output. Through the prompt text, the large language model can be explicitly told what task it is expected to perform, such as answering a question, simulating a conversation, writing an article, translating text, etc. At the same time, the prompt text can provide the necessary background information and context for the large language model to understand the logic, style, theme, or position that should be followed when generating content. In addition, the prompt text can also stimulate the large language model to exhibit its inherent knowledge reserves or specific language abilities, such as explaining complex concepts, citing regulations, imitating the writing style of a specific author, etc.

[0043] In order to improve the adaptability and response accuracy of the intelligent dialogue system, an external knowledge base can be used, so that the intelligent dialogue system, when answering the question raised by the user, no longer relies only on the limited knowledge obtained by the large language model in the training process through learning static corpus, but can first perform information retrieval and reasoning in the external knowledge base based on the question to obtain relevant information, and then understand and answer the question based on the relevant information, and generate the corresponding answer. That is, the external knowledge base can be combined with the large language model to obtain relevant information from the external knowledge base during model generation to assist the model to make more accurate and comprehensive answers or decisions. Since the relevant information obtained and the context of the question are considered during model generation, it can be ensured that the generated content is both in line with the actual needs and accurate, reliable, coherent and natural.

[0044] In the present application, since the intelligent dialogue system needs to provide a guide service to the user, it means that the question content raised by the user usually contains personal information such as gender and age, medical history, medication history, symptoms, diseases, test / check items, surgical operation, etc., and the user expects to get an answer to the department he can go to. Therefore, the relevant information obtained from the external knowledge base should be information related to the department, so as to better assist the large language model in the intelligent dialogue system to generate an answer containing the department.

[0045] One or more embodiments of the present application provide a technical solution for realizing intelligent guide, in which a guide graph can be constructed in advance, the nodes in the graph include nodes representing medical feature named entities and nodes representing department named entities, and the medical feature named entities represented by the nodes connected by an edge correspond to the department named entities represented by the nodes connected by the edge; subsequently, when the query text for guide consultation is obtained, named entity recognition can be performed on the query text to identify named entities and their types from the query text, if the identified named entities include disease feature named entities, the disease feature named entities can be input into a trained prediction model, the prediction model predicts the corresponding disease named entities based on the disease feature named entities, and based on the graph, the department named entities corresponding to the disease named entities are determined, if the identified named entities include medical feature named entities, the department named entities corresponding to the medical feature named entities can be directly determined based on the graph, and then the department named entities can be input into a guide large model, the guide large model performs reasoning based on the department named entities to generate an answer text corresponding to the query text, and the answer text can be used for guide.

[0046] By adopting the above manner, the intelligent guide diagnosis service is realized, so that the medical treatment process of the patient can be simplified and the efficiency is improved, the medical institution can help to relieve the pressure of human resources, and the service quality is optimized, thereby helping to improve the overall medical service experience. In addition, in the process of providing the intelligent guide diagnosis service, the trained prediction model and the constructed graph can be used to determine the corresponding department named entity based on various medical feature named entities in the query text for guide diagnosis consultation, and the determined department named entity is integrated into the real input of the guide diagnosis large model, so that the guide diagnosis large model can consider the internal generalization knowledge and the specific knowledge reflected by the department named entity when generating the answer text corresponding to the query text, thereby improving the adaptability and response accuracy of the guide diagnosis large model.

[0047] Please refer to Figure 1 , Figure 1 is a schematic diagram of an intelligent conversation system according to an example embodiment of the present application.

[0048] As shown in Figure 1 , the above intelligent conversation system can include a server and at least one client accessing the server through any type of wired or wireless network.

[0049] The above server can correspond to a server including a single independent physical host, or a server cluster composed of multiple independent physical hosts; or can correspond to a virtual server, a cloud server, etc. carried by a host cluster.

[0050] The above client can correspond to a terminal device such as a smart phone, a tablet computer, a notebook computer, a desktop computer, a PC (Personal Computer), a PDA (Personal Digital Assistant), a wearable device (for example: smart glasses, smart watch, etc.), a smart vehicle device or a game console, etc.

[0051] The user can use the intelligent conversation service provided by the above intelligent conversation system through the above client; the above client and the above server can realize the intelligent conversation service for the user through data interaction between each other.

[0052] For example, the client can output a corresponding user interface to the user, so that the user can perform operations such as inputting query text (which can be referred to as Query or Question), uploading documents or pictures for assisting in asking questions, etc. in the user interface to ask questions to the intelligent dialogue system using the intelligent dialogue service provided by the intelligent dialogue system. The client can send the query text input by the user to the server, generate a corresponding answer text (which can be referred to as Answer) for the query text by the server, and output the answer text to the user, that is, return the answer text to the client, and display the answer text to the user through the user interface, so that the user can view the corresponding answer generated by the intelligent dialogue system, thereby realizing the intelligent dialogue service for the user.

[0053] Specifically, the above-mentioned server can be loaded with a large language model, and the above-mentioned intelligent dialogue system can be based on the large language model, and the large language model can understand and answer the query text input by the user, and generate an answer text corresponding to the query text.

[0054] At this time, the above-mentioned large language model can refer to the service model of the large language model. In actual application, for the constructed large language model, an unsupervised learning method can be used to pre-train on a large-scale, unlabeled text data set to obtain a base model of the large language model; further, the dialogue task can be used as a supervised learning task during fine-tuning, and a text data set specific to the dialogue task can be prepared, so that the base model of the large language model can be used as a starting point for fine-tuning, and a supervised learning method can be used to fine-tune on the text data set specific to the dialogue task to obtain the service model of the large language model.

[0055] In order to improve the adaptability and response accuracy of the above-mentioned intelligent dialogue system, the above-mentioned server can also be loaded with a knowledge base and an information retrieval component. The knowledge base is an external knowledge base relative to the large language model loaded on the server, that is, the data in the external knowledge base is not knowledge learned by the large language model during training, but is auxiliary data in the reasoning process of the large language model to assist the large language model to generate an answer text corresponding to the query text input by the user. In the reasoning process of the large language model, the information retrieval component can perform information retrieval and reasoning in the external knowledge base based on the above-mentioned query text to assist the large language model to generate an answer text corresponding to the query text input by the user through the obtained related information. In addition, the above-mentioned server can also be loaded with a prediction model for simplifying information retrieval and reasoning in the external knowledge base to a certain extent.

[0056] In practical applications, the server end can also carry other functional components or functional subsystems such as a prompt generation component. These components or subsystems can work in coordination with the large language model carried on the server end to jointly generate an answer text corresponding to the query text input by the user.

[0057] Please refer to Figure 2 , Figure 2 is a schematic diagram of an intelligent guide diagnosis process according to an example embodiment of the present application.

[0058] As Figure 2 shown, in the above intelligent guide diagnosis process, an external knowledge base for assisting model generation can be pre-constructed, which can be a graph.

[0059] A graph is data stored and managed in a graph structure. In a graph, nodes, edges, and properties are used to store data. This storage method is very suitable for expressing complex relationships between entities. In a graph, nodes represent entities such as people, places, events, etc., and each node can have multiple properties to describe the specific information of the entity. Edges are used to represent the relationship between nodes, such as "knows", "belongs to", "located in", etc., and edges can also contain properties to describe the characteristics of the relationship, such as the strength of the relationship, the establishment time, etc. Properties are data fields attached to nodes or edges to store specific information, such as a person's name, age, or relationship start date, etc.

[0060] For intelligent guide diagnosis services, the above graph can represent the correspondence between diseases, examination / inspection items, surgical operations, etc. and departments, i.e. where to go for treatment when suffering from a certain disease, needing to do a certain examination / inspection item, or needing to do a certain surgical operation. In this application, these information that can be used to determine the treatment department is called a treatment feature. Specifically, in the graph, nodes can be divided into treatment feature nodes and department nodes, a treatment feature node can represent a treatment feature named entity (i.e. the identifying text of the treatment feature, for example: disease name text), and a department node can represent a department named entity (i.e. the identifying text of the department, for example: department name text); an edge is used to connect a treatment feature node and a department node, where the treatment feature named entity represented by the treatment feature node corresponds to the department named entity represented by the department node (i.e. when having the treatment feature, the department needs to be treated); the treatment feature named entity can include at least one sub-type of named entity, since disease is usually the decisive factor for the treatment department, the at least one sub-type can include disease, and in addition, examination / inspection items, surgical operations, etc. can also be included according to actual situation and demand.

[0061] It should be noted that, since one disease can have multiple disease characteristics, and the same disease characteristic can be caused by different diseases, a machine learning method can be used to obtain a prediction model for predicting the corresponding disease based on the disease characteristics through model training. In this case, the prediction model can be used to predict the corresponding disease based on the disease characteristics first, and then the information retrieval and reasoning based on the disease in the above diagram can be performed to determine the corresponding department. Among them, the disease characteristics can not only include symptoms closely related to the disease (for example: frequent urination, urgent urination, dysuria, urinary pain, and lumbar pain), but also include gender (for example: male, female), age, and other information that can affect the diagnosis of the disease to some extent. For example, men have a prostate, while women do not have a prostate, so a male patient with the symptoms of frequent urination and urgent urination may be diagnosed as having benign prostatic hyperplasia, but a female patient with the symptoms of frequent urination and urgent urination cannot be diagnosed as having benign prostatic hyperplasia; that is, gender can affect the diagnosis of benign prostatic hyperplasia. In addition, according to actual needs, medical history, medication history, and other information can also be used as disease characteristics, which are not specially limited in the present application.

[0062] In order to reduce the training amount required to obtain the guidance large model, and reduce the difficulty and complexity of obtaining the guidance large model, knowledge distillation (KD) can be used to extract guidance-related knowledge from the service model of the large language model that can provide more comprehensive services, and compress the knowledge into the base model of another large language model, so that the base model becomes the service model of the large language model that can provide intelligent guidance services, and the service model of the large language model obtained at this time can be used as the guidance large model. Specifically, knowledge distillation can be performed on a preset service large model (i.e., the service model of the large language model that can provide more comprehensive services, such as ChatGPT) as a teacher model to extract guidance-related knowledge from the service large model, and compress the knowledge into a large language model as a student model to obtain the guidance large model. When performing knowledge distillation on the service large model, the training data used can be a query text dataset for guidance consultation, and the label can be a guidance answer text corresponding to the query text.

[0063] In the process of intelligent guidance, first, the query text for guidance consultation can be obtained, and named entity recognition is performed on the query text to identify named entities from the query text and determine the types of the identified named entities; if the identified named entities include disease feature named entities, the disease feature named entities can be input into the trained prediction model, and the prediction model predicts the corresponding disease named entities based on the disease feature named entities, and then determines the department named entities corresponding to the disease named entities based on the above graph; if the identified named entities include consultation feature named entities, the department named entities corresponding to the consultation feature named entities can be directly determined based on the above graph; finally, the department named entity can be input into the above guidance large model, and the guidance large model can generate an answer text corresponding to the query text based on the department named entity, which can be used for guidance.

[0064] It should be noted that the above graph can be constructed, the above prediction model can be trained, and the above guidance large model can be trained offline, and the intelligent guidance using the constructed graph, the trained prediction model, and the knowledge distilled guidance large model can be completed online.

[0065] The intelligent guidance method provided by the present application will be described in detail from the following four aspects of graph construction, prediction model training, guidance large model obtaining, and intelligent guidance.

[0066] I. Graph construction

[0067] First of all, it should be noted that in the constructed graph, nodes can be divided into consultation feature nodes and department nodes, a consultation feature node can represent a consultation feature named entity (i.e., an identifying text of a consultation feature, such as a disease name text), and a department node can represent a department named entity (i.e., an identifying text of a department, such as a department name text); an edge is used to connect a consultation feature node and a department node, wherein the consultation feature named entity represented by the consultation feature node corresponds to the department named entity represented by the department node (i.e., the consultation feature needs to be treated in the department); the consultation feature named entity can include at least one sub-type of named entity, and since disease is usually a decisive factor for consultation department, the at least one sub-type can include disease.

[0068] In some embodiments, the above graph can be constructed according to a method of constructing a knowledge graph. That is, the constructed above graph can be a knowledge graph.

[0069] In some embodiments, the at least one sub-type can further include one or more of the following according to actual conditions and needs: a medical examination item; a medical examination item; a medical operation. Among them, the medical examination generally refers to the process of analyzing the collected samples (such as blood, urine, tissue, etc.) by the laboratory to determine the composition, nature or state thereof, and the medical examination refers to the physical examination, imaging examination (such as X-ray, CT, MRI, etc.), endoscopy and other non-laboratory operations performed by doctors or other medical professionals on patients.

[0070] In the process of constructing the above graph, medical samples can be obtained, and named entity recognition can be performed on each medical sample to identify corresponding medical feature named entities and department named entities from the medical samples. Among them, the medical samples usually record the diseases diagnosed by the patient, the examination / inspection items and operation recommended, etc., and also record the department where the patient is treated; for example, the medical samples can be open source and desensitized medical records.

[0071] For a medical sample, since a medical sample usually records the diseases diagnosed by a patient, the examination / inspection items and operation recommended, etc., and also records the department where the patient is treated, the medical feature named entities and department named entities obtained by performing named entity recognition on the medical sample are corresponding medical feature named entities and department named entities.

[0072] In some embodiments, a large language model can be used for convenient named entity recognition. Specifically, each medical sample can be input into the large language model, and the large language model can perform named entity recognition on each medical sample to identify corresponding medical feature named entities and department named entities from the medical samples.

[0073] The large language model at this time can refer to a service model of the large language model. In actual application, for the constructed large language model, an unsupervised learning manner can be adopted to pre-train on a large-scale, unlabeled text data set to obtain a base model of the large language model; further, the named entity recognition task can be used as a supervised learning task during fine-tuning, and a text data set specific to the named entity recognition task can be prepared, so that the base model of the large language model can be used as a starting point for fine-tuning, and a supervised learning manner can be adopted to fine-tune on the text data set specific to the named entity recognition task to obtain the service model of the large language model.

[0074] In some embodiments, in order to ensure the consistency of all named entity representations, the named entities identified by the named entity recognition can be standardized. For example, "cardiology department", "cardiovascular internal medicine department", and "cardiology department" can be standardized to a standard form, i.e., "cardiovascular internal medicine department".

[0075] In the case of the identified above corresponding medical feature named entity and department named entity, the above graph can be constructed according to the corresponding medical feature named entity and department named entity.

[0076] For example, assuming that medical feature named entity A, medical feature named entity B, and department named entity M are identified for medical sample 1, and medical feature named entity C and department named entity N are identified for medical sample 2, it can be known that medical feature named entity A and medical feature named entity B correspond to department named entity M respectively, and medical feature named entity C corresponds to department named entity N, so the constructed graph can be as shown in Figure 3A , where node A represents medical feature named entity A, node B represents medical feature named entity B, node C represents medical feature named entity C, node M represents department named entity M, and node N represents department named entity N. Node A and node M are connected by an edge, node B and node M are connected by an edge, and node C and node N are connected by an edge.

[0077] In some embodiments, in order to make the relationship between different entities more clear and explicit, and to facilitate the subsequent use of the above graph to determine the department corresponding to a specific medical feature, a corresponding weight can be set for each edge in the graph. Specifically, for an edge connecting a medical feature node and a department node, the weight of this edge can be the association degree of the medical feature named entity represented by this medical feature node relative to the department named entity represented by this department node.

[0078] In the above case, for various subtypes (which can be referred to as target subtypes) of the named entity of the medical feature named entity identified from the medical sample, the named entity of the target subtype can be taken as a training sample, and the department named entity corresponding to the named entity of the target subtype can be taken as a label of the training sample. Based on the training sample and the label thereof, a preset prediction model is supervised trained to obtain a trained prediction model. Then, each named entity of the target subtype is input into the trained prediction model, and the trained prediction model predicts the association degree (which can be specifically a probability of the prediction model predicting each department named entity based on each named entity of the target subtype) of each named entity of the target subtype relative to each department named entity. Subsequently, the medical feature named entity and the department named entity with the association degree greater than a preset threshold can be determined as the corresponding medical feature named entity and department named entity, and the graph can be constructed according to the identified medical feature named entity and department named entity, and the corresponding medical feature named entity and department named entity and the association degree therebetween.

[0079] It should be noted that the prediction models for predicting the association degrees of different subtypes of the named entity relative to each department named entity are usually different prediction models.

[0080] For example, assuming that the medical feature named entity A, the medical feature named entity B and the department named entity M are identified for the medical sample 1, the medical feature named entity C and the department named entity N are identified for the medical sample 2, and the trained prediction model predicts that the association degree of the medical feature named entity A relative to the department named entity M is 0.9, the association degree of the medical feature named entity B relative to the department named entity M is 0.78, and the association degree of the medical feature named entity C relative to the department named entity N is 0.88, and the preset threshold is 0.8, it can be known that the medical feature named entity A corresponds to the department named entity M and the association degree therebetween is 0.9, and the medical feature named entity C corresponds to the department named entity N and the association degree therebetween is 0.88. Therefore, the constructed graph can be as shown in FIG. 1, where the node A represents the medical feature named entity A, the node B represents the medical feature named entity B, the node C represents the medical feature named entity C, the node M represents the department named entity M, and the node N represents the department named entity N. The node A and the node M are connected by an edge, and the weight of the edge is 0.9. The node C and the node N are connected by an edge, and the weight of the edge is 0.88. Figure 3B

[0081] II. Training of the prediction model

[0082] ​In the process of training the above prediction model, medical samples can be obtained, and named entity recognition can be performed on each medical sample to identify corresponding disease feature named entities and disease named entities from the medical samples. Among them, the medical samples usually record the personal information, symptoms, etc. of the patient, and also record the disease diagnosed by the patient; for example, the medical samples can be open source and desensitized medical records.

[0083] For a medical sample, since a medical sample usually records the personal information, symptoms, etc. of a patient, and also records the disease diagnosed by the patient, the disease feature named entities and disease named entities obtained by performing named entity recognition on the medical sample are the corresponding disease feature named entities and disease named entities.

[0084] In some embodiments, a large language model can be used for convenient named entity recognition. Specifically, each medical sample can be input into the large language model, and the large language model can perform named entity recognition on each medical sample to identify corresponding disease feature named entities and disease named entities from the medical samples.

[0085] At this time, the above large language model can refer to the service model of the large language model. It should be noted that in order to ensure the consistency of the identified named entities, the large language model and the large language model used for named entity recognition in the construction process of the above diagram can be the same large language model.

[0086] In some embodiments, in order to ensure the consistency of all named entities, the named entities identified by named entity recognition can be standardized. For example, “cardiology department”, “cardiovascular internal medicine department”, and “cardiology department” can be standardized to a standard form, i.e. “cardiovascular internal medicine department”.

[0087] After identifying the above corresponding disease feature named entities and disease named entities, the disease feature named entities can be used as training samples, and the disease named entities corresponding to the disease feature named entities can be used as labels of the training samples. Based on the training samples and their labels, the pre-set prediction model is supervised trained to obtain the trained prediction model.

[0088] It should be noted that the prediction model for predicting the corresponding department named entity based on the visit feature named entity is usually different from the prediction model for predicting the corresponding disease named entity based on the disease feature named entity.

[0089] In some embodiments, the above prediction model for predicting a corresponding disease named entity based on a disease feature named entity can be specifically a decision tree model.

[0090] III. Obtaining the Guiding Diagnosis Large Model

[0091] To reduce the amount of training required to obtain the guiding diagnosis large model, and to reduce the difficulty and complexity of obtaining the guiding diagnosis large model, knowledge distillation (KD) can be used to extract guiding diagnosis-related knowledge from a large language model service model that can provide more comprehensive services, and compress the knowledge into a base model of another large language model, so that the base model becomes a large language model service model that can provide intelligent guiding diagnosis services. At this time, the large language model service model obtained can be used as the guiding diagnosis large model.

[0092] Knowledge distillation is a machine learning technique mainly used to reduce the complexity of deep learning models and improve the efficiency of the models. The basic idea of knowledge distillation is to extract knowledge from a complex, large model (called a teacher model) and "compress" it into a smaller, simpler model (called a student model). The essence of knowledge distillation lies in knowledge extraction, knowledge transfer, and knowledge compression; among them, knowledge extraction refers to extracting effective decision information from the teacher model, knowledge transfer refers to transferring this information to the student model in a learnable form, and knowledge compression refers to ensuring that the student model can reproduce the behavior of the teacher model with a smaller scale.

[0093] Knowledge distillation mainly includes the following steps: first, train a teacher model with excellent performance but may be more complex. Use the teacher model to predict the training data to obtain soft labels (Soft Labels); among them, the soft label is usually a probability distribution after temperature scaling, which can provide more information than the hard label (Hard Labels). Then train a smaller student model to make its prediction as close as possible to the soft label of the teacher model. During this process, hard labels and soft labels may be used as part of the loss function. Finally, evaluate the performance of the student model and make necessary adjustments to optimize its performance.

[0094] That is, a loss function can be defined to measure the difference between the prediction results of the student model and the prediction results of the teacher model, and soft targets are usually used to achieve this, that is, the output of the teacher model is used as the target of the student model, and the cross-entropy loss or other similar loss functions are calculated. Subsequently, the prediction results of the teacher model can be used as an auxiliary target, combined with the original target of the student model, to train the student model by iteratively minimizing the loss function to update the model parameters of the student model.

[0095] Specifically, knowledge distillation can be performed on a preset service large model (i.e., a service model of a large language model capable of providing more comprehensive services, such as ChatGPT) as a teacher model to extract knowledge related to the guide from the service large model and compress the knowledge into a large language model as a student model to obtain the guide large model. When performing knowledge distillation on the service large model, the training data used can be a query text dataset for guide consultation, and the label can be a guide answer text corresponding to the query text.

[0096] IV. Intelligent guide

[0097] Please refer to Figure 4 , Figure 4 is a flowchart of an intelligent guide method according to an example embodiment of the present application.

[0098] As shown in Figure 4 , the intelligent guide method can include the following steps:

[0099] Step 402: Obtain a query text for guide consultation, and perform named entity recognition on the query text to identify named entities from the query text and determine the types of the identified named entities.

[0100] In this embodiment, a query text for guide consultation can be obtained. After obtaining the query text, named entity recognition can be performed on the query text to identify named entities from the query text and determine the types of the identified named entities.

[0101] In some embodiments, a large language model can be used to perform named entity recognition on the query text. Specifically, the query text can be input into the large language model, and the large language model can perform named entity recognition on the query text.

[0102] The large language model at this time can refer to the service model of the large language model. It should be noted that in order to ensure the consistency of the identified named entities, the large language model can be the same as the large language model used for named entity recognition in the process of constructing the above graph and training the above prediction model.

[0103] In some embodiments, in order to ensure consistency of all named entity representations, the named entities identified by the named entity recognition can be standardized. For example, "cardiology department", "cardiovascular internal medicine department", and "cardiology department" can be standardized to a standard form, i.e., "cardiovascular internal medicine department". That is, the named entity recognition can be performed on the above query text, and the identified named entities can be standardized to determine the standardized named entities as the named entities identified from the query text, and determine the types of the identified named entities.

[0104] Step 404: If the identified named entity includes a disease feature named entity, input the disease feature named entity into the trained prediction model, predict the corresponding disease named entity based on the disease feature named entity by the prediction model, and determine the department named entity corresponding to the disease named entity based on the graph.

[0105] In this embodiment, if it is determined according to the type of the named entity identified from the above query text that the identified named entity includes a disease feature named entity, the disease feature named entity can be first input into the trained prediction model, the corresponding disease named entity can be predicted based on the disease feature named entity by the prediction model, and then the department named entity corresponding to the disease named entity can be determined based on the constructed graph.

[0106] In some embodiments, the disease named entity predicted by the prediction model is usually unique, in which case the correlation degree of the disease named entity relative to each department named entity can be calculated based on the graph, and the department named entity with the highest correlation degree is determined as the department named entity corresponding to the disease named entity. Specifically, if the edges in the graph have no weights, one department named entity can be randomly selected from all department named entities represented by department nodes connected to the disease node representing the disease named entity as the department named entity with the highest correlation degree to the disease named entity; if the edges in the graph have weights, the edge with the highest weight can be selected from all edges connecting the disease node representing the disease named entity, and the department named entity represented by the department node connected by the edge is determined as the department named entity with the highest correlation degree to the disease named entity.

[0107] Step 406: If the identified named entity includes the medical feature named entity, determine the department named entity corresponding to the medical feature named entity based on the graph.

[0108] In the embodiment, if it is determined that the identified named entity includes a medical feature named entity according to the type of the named entity identified from the query text, the department named entity corresponding to the medical feature named entity can be determined based on the constructed graph.

[0109] In some embodiments, the association degree of the medical feature named entity with respect to each department named entity can be calculated based on the graph, and the department named entity with the highest association degree can be determined as the department named entity corresponding to the medical feature named entity. Specifically, if the edges in the graph have no weights, the department named entity represented by the department node connected to the medical feature node representing the medical feature named entity through the edge with the largest number of connected edges can be selected as the department named entity with the highest association degree with the medical feature named entity; if the edges in the graph have weights, the department named entity represented by the department node connected to the medical feature node representing the medical feature named entity through the edge with the largest sum of weights of connected edges can be selected as the department named entity with the highest association degree with the medical feature named entity.

[0110] For example, assuming that the medical feature named entity A, the medical feature named entity B, and the medical feature named entity C are identified for the query text, the node A representing the medical feature named entity A and the node M representing the department entity M are connected by an edge, the node B representing the medical feature named entity B and the node M are connected by an edge, the node C representing the medical feature named entity C and the node M are connected by an edge, the node C and the node N representing the department entity N are connected by an edge, and the edges have no weights, since the number of edges connected to the node M is 3 and the number of edges connected to the node N is 1, the department entity M represented by the node M can be determined as the department named entity with the highest association degree with the medical feature named entity.

[0111] For example, assume that the clinic feature named entity A, the clinic feature named entity B, and the clinic feature named entity C are identified for the above query text, where a node A representing the clinic feature named entity A and a node M representing the department entity M are connected by an edge with a weight of 0.9, a node B representing the clinic feature named entity B and the node M are connected by an edge with a weight of 0.78, the node B and a node N representing the department entity N are connected by an edge with a weight of 0.82, and a node C representing the clinic feature named entity C and the node N are connected by an edge with a weight of 0.88. Since the sum of the weights of the edges connected to the node M is 0.9 + 0.78 = 1.68, and the sum of the weights of the edges connected to the node N is 0.82 + 0.88 = 1.7, the department entity N represented by the node N can be determined as the department named entity with the highest correlation degree to the clinic feature named entity.

[0112] Step 408: inputting the department named entity into the guidance large model, and performing reasoning by the guidance large model based on the department named entity to generate an answer text corresponding to the query text.

[0113] In this embodiment, after the above department named entity is determined, the department named entity can be input into the above guidance large model to perform reasoning by the guidance large model based on the department named entity to generate an answer text corresponding to the query text (usually, the answer text contains the department named entity), which can be used for guidance. For example, based on the department named entity, a prompt text for prompting the guidance large model to perform a guidance task (which can be regarded as a dialogue task in nature) using the department named entity can be constructed, and the prompt text can be input into the guidance large model to perform reasoning by the guidance large model based on the department named entity under the guidance of the prompt text to generate an answer text corresponding to the query text.

[0114] In the technical solution, a graph for guiding diagnosis can be constructed in advance, nodes in the graph include nodes representing appointment feature named entities and nodes representing department named entities, and an appointment feature named entity represented by a node connected by an edge corresponds to a department named entity represented by a node connected by the edge. Subsequently, when a query text for guiding diagnosis consultation is obtained, named entity recognition can be performed on the query text to identify named entities and their types from the query text. If the identified named entities include disease feature named entities, the disease feature named entities can be input into a trained prediction model first, and the prediction model can predict corresponding disease named entities based on the disease feature named entities. Then, based on the graph, a department named entity corresponding to the disease named entity can be determined. If the identified named entities include appointment feature named entities, a department named entity corresponding to the appointment feature named entity can be directly determined based on the graph. The department named entity can be input into a guiding diagnosis large model, and the guiding diagnosis large model can perform reasoning based on the department named entity to generate an answer text corresponding to the query text. The answer text can be used for guiding diagnosis.

[0115] In the foregoing manner, intelligent guiding diagnosis services are achieved, thus simplifying the medical treatment process of patients and improving efficiency, helping medical institutions to relieve the pressure of human resources and optimize service quality, thereby helping to improve the overall medical service experience. In addition, in the process of providing intelligent guiding diagnosis services, the trained prediction model and the constructed graph can be used to determine corresponding department named entities based on various appointment feature named entities in a query text for guiding diagnosis consultation, and the determined department named entities can be integrated into the real input of the guiding diagnosis large model. When the guiding diagnosis large model generates an answer text corresponding to the query text, it can consider both the inherent generalization knowledge and the specific knowledge reflected by the department named entity, thereby improving the adaptability and response accuracy of the guiding diagnosis large model.

[0116] Corresponding to the embodiments of the foregoing method, the present application also provides embodiments of an apparatus.

[0117] For reference Figure 5 , Figure 5is a structural schematic diagram of a device shown in an exemplary embodiment of the present application. At the hardware level, the device includes a processor 502, an internal bus 504, a network interface 506, a memory 508, and a non-volatile memory 510, and of course can also include other required hardware. One or more embodiments of the present application can be implemented in a software manner, such as reading a corresponding computer program from the non-volatile memory 510 into the memory 508 by the processor 502 and then running. Of course, in addition to the software implementation, one or more embodiments of the present application do not exclude other implementation manners, such as a logic device or a combination of software and hardware, and the like, that is, the execution subject of the following processing flow is not limited to each logic module, but can also be hardware or a logic device.

[0118] Please refer to Figure 6 , Figure 6 is a block diagram of an intelligent guide diagnosis device shown in an exemplary embodiment of the present application.

[0119] The above intelligent guide diagnosis device can be applied to Figure 5 the device shown in the figure to implement the technical solutions of the present application. The device includes:

[0120] The acquisition module 602 acquires a query text for guide diagnosis consultation, and performs named entity recognition on the query text to identify named entities from the query text and determine the types of the identified named entities;

[0121] The first determination module 604 inputs the disease feature named entity into the trained prediction model if the identified named entity includes the disease feature named entity, predicts the corresponding disease named entity based on the disease feature named entity by the prediction model, and determines the department named entity corresponding to the disease named entity based on a graph. In the graph, the nodes include a medical feature node and a department node, each medical feature node represents each medical feature named entity, and each department node represents each department named entity. The edges in the graph are used to connect the medical feature nodes and the department nodes, and the medical feature named entity represented by the medical feature node connected by each edge corresponds to the department named entity represented by the department node connected by the edge. The medical feature named entity includes at least one sub-type of named entity, and the at least one sub-type includes a disease.

[0122] The second determination module 606 determines the department named entity corresponding to the medical feature named entity based on the graph if the identified named entity includes the medical feature named entity.

[0123] The reasoning module 608 inputs the department named entity into the guide large model, and the guide large model performs reasoning based on the department named entity to generate answer text corresponding to the query text.

[0124] In some embodiments, the device further comprises a prediction model training module for:

[0125] Obtaining a medical sample, and performing named entity recognition on the medical sample to identify corresponding disease feature named entities and disease named entities from the medical sample;

[0126] Taking the disease feature named entity as a training sample, and taking the disease named entity corresponding to the disease feature named entity as a label of the training sample, based on the training sample and its label, performing supervised training on the prediction model.

[0127] In some embodiments, the prediction model is a decision tree model.

[0128] In some embodiments, the device further comprises a graph construction module for:

[0129] Obtaining a medical sample, and performing named entity recognition on the medical sample to identify corresponding disease feature named entities and disease named entities from the medical sample;

[0130] According to the identified corresponding disease feature named entities and department named entities, constructing the graph.

[0131] In some embodiments, each edge in the graph is provided with a corresponding weight, and the weight of each edge is the correlation degree of the disease feature named entity represented by the disease feature node connected by the edge relative to the department named entity represented by the department node connected by the edge.

[0132] The device further comprises a correlation degree calculation module for:

[0133] Taking various target sub-type named entities in the identified disease feature named entities as training samples, and taking the department named entities corresponding to the target sub-type named entities as labels of the training samples, based on the training samples and their labels, performing supervised training on a preset prediction model;

[0134] Inputting each named entity of the target sub-type into the trained prediction model, and predicting the correlation degree of each named entity of the target sub-type relative to each department named entity by the prediction model;

[0135] The constructing the graph according to the identified corresponding disease feature named entities and department named entities comprises:

[0136] determine the appointment feature named entity and the department named entity with the correlation greater than the preset threshold as the corresponding appointment feature named entity and the department named entity, and construct the graph according to the identified appointment feature named entity and the department named entity, and the corresponding appointment feature named entity and the department named entity.

[0137] In some embodiments, the determining, based on the graph, the department named entity corresponding to the named entity comprises:

[0138] Based on the graph, the correlation of the named entity with respect to each department named entity is calculated, and the department named entity with the greatest correlation is determined as the department named entity corresponding to the named entity.

[0139] In some embodiments, the at least one sub-type further comprises one or more of the following: a medical examination item; a medical examination item; a medical operation.

[0140] In some embodiments, the device further comprises a knowledge distillation module for:

[0141] Knowledge distillation is performed on a preset service large model as a teacher model to extract knowledge related to the guidance from the service large model, and the knowledge is compressed into a large language model as a student model to obtain the guidance large model.

[0142] In some embodiments, the performing named entity recognition on the query text to identify named entities from the query text and determining the type of each identified named entity comprises:

[0143] Named entity recognition is performed on the query text, and standardized processing is performed on the identified named entities to determine the standardized named entities as the named entities identified from the query text, and determine the type of each identified named entity.

[0144] For the device embodiment, it basically corresponds to the method embodiment, so for the related part, please refer to the part of the method embodiment. The device embodiments described above are only illustrative, and the modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical modules, i.e. they can be located in one place or distributed on multiple network modules. According to actual needs, some or all of the modules can be selected to achieve the purpose of the technical solution of the present application.

[0145] The systems, apparatuses, modules, or components illustrated in the above examples can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer, and the specific form of the computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an e-mail device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0146] In one typical configuration, a computer includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0147] The memory can include non-persistent memory in the computer readable medium, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory. The memory is an example of computer readable media.

[0148] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented by any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, disk storage, quantum memory, graphene-based storage medium or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory computer readable media, such as modulated data signals and carriers.

[0149] It should be noted that the terms "comprising", "including", or any other variant thereof are intended to cover non-exclusive inclusion, so that processes, methods, articles or devices including a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "including a" does not exclude the presence of additional identical elements in the process, method, article or device including the element.

[0150] The above-described embodiments of the application have several aspects, no single one of which is solely responsible for the application's desirable attributes. Without limiting the scope of this application, several non-limiting examples of aspects of the application are discussed below.

[0151] The terminology used in this description is for the purpose of describing particular embodiments only and is not intended to limit one or more embodiments of the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. The term "and / or" includes any and all combinations of one or more of the associated listed items.

[0152] The terms "one embodiment," "an embodiment," "some embodiments," "exemplary" or "example" as may be used herein in the specification and / or claims mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment of the present application. The appearance of the phrases "in one embodiment" or "in some embodiments" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily referring to a single, "preferred" embodiment. Furthermore, the described features, advantages, characteristics, etc. of the application can be combined in any suitable manner in one or more embodiments.

[0153] It is to be understood that the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Pronouns in the masculine form include the feminine form, and vice versa, and the singular form also includes the plural form, unless the context clearly dictates otherwise. The terms "approximately" and "substantially" and variations thereof as used herein represent an allowance for measurement of natural variations in materials. These variations are unavoidable during manufacturing operations and therefore the cost of precise manufacturing procedures is not required. These variations can include, but are not limited to, variations that render the application and its uses as described herein more efficient, more effective, and / or otherwise improved. The use of the terms "if", as can be used herein, can be interpreted as meaning "when" or "upon" or "in response to determining" taking into account the context in which such term is used.

[0154] The foregoing is considered as illustrative only of the principles of the application. Numerous modifications and changes can be devised by those skilled in the art without departing from the spirit and principles of the application. The disclosure is not intended to limit the application to the precisely as described hereinabove.

[0155] The user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are information and data authorized by the user or authorized by all parties, and the collection, use and processing of the related data need to comply with the relevant laws, regulations and standards of the country and region, and provide corresponding operation entrances for the user to choose authorization or refusal.

Claims

1. An intelligent triage method, the method comprising: Obtain the query text used for triage consultation, and perform named entity recognition on the query text to identify named entities from the query text and determine the type of each identified named entity; If the identified named entities include disease feature named entities, then the disease feature named entities are input into the trained prediction model. The prediction model predicts the corresponding disease named entities based on the disease feature named entities, and determines the department named entities corresponding to the disease named entities based on the graph. The nodes in the graph include visit feature nodes and department nodes. Each visit feature node represents a visit feature named entity, and each department node represents a department named entity. The edges in the graph connect the visit feature nodes and the department nodes. The visit feature named entity represented by each edge corresponds to the department named entity represented by the department node connected by the edge. The visit feature named entities include at least one subtype of named entities, and the at least one subtype includes diseases. If the identified named entity includes the patient visit feature named entity, then based on the graph, determine the department named entity corresponding to the patient visit feature named entity; The named entity of the department is input into the patient guidance model, which then performs reasoning based on the named entity of the department to generate the answer text corresponding to the query text.

2. The method according to claim 1, further comprising: Acquire medical samples and perform named entity recognition on the medical samples to identify corresponding disease feature named entities and disease named entities from the medical samples; The disease feature named entities are used as training samples, and the disease named entities corresponding to the disease feature named entities are used as labels for the training samples. Based on the training samples and their labels, the prediction model is trained in a supervised manner.

3. The method according to claim 2, wherein the prediction model is a decision tree model.

4. The method according to claim 1, further comprising: Acquire medical samples and perform named entity recognition on the medical samples to identify corresponding medical visit feature named entities and department named entities from the medical samples; The graph is constructed based on the identified corresponding named entities for patient visit features and department names.

5. According to the method of claim 4, each edge in the figure is assigned a corresponding weight, and the weight of each edge is the degree of correlation between the named entity of the patient visit feature node connected by each edge and the named entity of the department node connected by the edge. The method further includes: The named entities of various target subtypes in the identified medical visit feature named entities are used as training samples, and the department named entities corresponding to the named entities of the target subtypes are used as labels for the training samples. Based on the training samples and their labels, supervised training is performed on the preset prediction model. Each named entity of the target subtype is input into the trained prediction model, and the prediction model predicts the correlation degree of each named entity of the target subtype with respect to the named entities of each department. The process of constructing the graph based on the identified corresponding patient visit feature named entities and department named entities includes: The named entities of patient visit features and departments with a correlation greater than a preset threshold are identified as the corresponding named entities of patient visit features and departments. The graph is constructed based on the identified named entities of patient visit features and departments, as well as the corresponding named entities of patient visit features and departments.

6. The method according to claim 1, wherein determining the department named entity corresponding to the named entity based on the graph includes: Based on the graph, the correlation degree of the named entity with respect to each department named entity is calculated, and the department named entity with the highest correlation degree is determined as the department named entity corresponding to the named entity.

7. The method according to claim 1, wherein the at least one subtype further includes one or more of the following: medical testing items; medical examination items; medical surgical procedures.

8. The method according to claim 1, further comprising: Knowledge distillation is performed on the pre-defined service model, which serves as the teacher model, to extract knowledge related to patient guidance. This knowledge is then compressed into a large language model, which serves as the student model, to obtain the patient guidance model.

9. The method according to claim 1, wherein performing named entity recognition on the query text to identify named entities from the query text and determining the type of each identified named entity includes: Named entity recognition is performed on the query text, and the identified named entities are standardized to determine the standardized named entities as those identified from the query text, and the type of each identified named entity is determined.

10. An intelligent triage device, the device comprising: The acquisition module acquires the query text used for triage consultation, and performs named entity recognition on the query text to identify named entities from the query text and determine the type of each identified named entity. The first determining module, if the identified named entities include disease feature named entities, inputs the disease feature named entities into the trained prediction model. The prediction model predicts the corresponding disease named entities based on the disease feature named entities, and determines the department named entities corresponding to the disease named entities based on the graph. The nodes in the graph include visit feature nodes and department nodes; each visit feature node represents a visit feature named entity, and each department node represents a department named entity. The edges in the graph connect the visit feature nodes and the department nodes; the visit feature named entity represented by each edge corresponds to the department named entity represented by the department node connected by the edge. The visit feature named entities include at least one subtype of named entities, and the at least one subtype includes diseases. The second determining module, if the identified named entity includes the medical visit feature named entity, then determines the department named entity corresponding to the medical visit feature named entity based on the graph; The reasoning module inputs the named entity of the department into the patient guidance model, and the patient guidance model performs reasoning based on the named entity of the department to generate the answer text corresponding to the query text.

11. An electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor implements the method as described in any one of claims 1 to 9 by executing the executable instructions.

12. A computer-readable storage medium having stored thereon computer instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 9.

Images