Page information push method and intelligent nursing system applied to user health management

Abstract

The embodiment of the present application provides a page information push method and a smart nursing system for user health management, which generates first knowledge node path data by integrating knowledge node data of the same nursing migration stage, integrates the first knowledge node path data of the same nursing knowledge theme again, forms second knowledge node path data, and uses a graph neural network model to process the second knowledge node path data to generate a first graph coding feature, and identifies the key knowledge node path data that best matches the user's needs by calculating the feature distance between the personalized nursing knowledge point data and the first graph coding feature of each second knowledge node path data. Finally, for the subsequent nursing stage corresponding to the key knowledge node path data, it is possible to push nursing execution suggestion information that is highly matched with the personalized nursing knowledge point data, thereby realizing personalized customization and accurate push of nursing suggestions, and significantly improving the efficiency and effectiveness of user health management.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a page information push method and an intelligent nursing system applied to user health management. Background Art

[0002] In the current health management field, with the rapid development of big data and artificial intelligence technology, the push of personalized care suggestions has become an important means to improve the user's health management level. However, related technologies often rely on simple questionnaires or manual analysis, which makes it difficult to fully and accurately capture the complex relationship between users' health needs and care knowledge.

[0003] For example, user health care knowledge is usually scattered in a large amount of data, which may come from different medical records, health monitoring equipment, professional literature, etc. These data are not only huge in quantity but also complex in structure. Traditional solutions often find it difficult to effectively integrate and utilize this information, resulting in the lack of accuracy and comprehensiveness in the push of personalized care recommendations. Summary of the invention

[0004] In order to at least overcome the above-mentioned deficiencies in the prior art, the purpose of the embodiments of the present application is to provide a page information push method and a smart care system for user health management.

[0005] According to one aspect of the present application, a page information push method for user health management is provided, the method comprising:

[0006] Integrate and output the knowledge node data corresponding to the same nursing migration stage in the user nursing health knowledge graph data as first knowledge node path data, wherein the user nursing health knowledge graph data is a data set to be analyzed and personalized nursing suggestions to be pushed;

[0007] Integrate and output the first knowledge node path data corresponding to the same nursing knowledge topic in the user nursing health knowledge graph data as second knowledge node path data;

[0008] Loading the second knowledge node path data into a graph neural network model to generate a first graph encoding feature for expressing the second knowledge node path data;

[0009] Acquire personalized nursing knowledge point data, determine a first feature distance between the personalized nursing knowledge point data and the first graph encoding feature of each of the second knowledge node path data, and determine key knowledge node path data among a plurality of the second knowledge node path data based on the first feature distance;

[0010] Nursing execution suggestion information matching the personalized nursing knowledge point data is pushed for the subsequent nursing stage corresponding to the key knowledge node path data.

[0011] According to one aspect of the present application, a smart care system is provided, which includes a processor and a machine-readable storage medium, wherein the machine-readable storage medium stores machine-executable instructions, and the machine-executable instructions are loaded and executed by the processor to implement the above-mentioned page information push method applied to user health management.

[0012] In the technical solutions provided in some embodiments of the present application, the embodiments of the present application realize the accurate push of personalized nursing suggestions by integrating the relevant information in the user's nursing health knowledge graph data. Specifically, the knowledge node data is first integrated into the first knowledge node path data according to the nursing migration stage, and then the first knowledge node path data is further integrated into the second knowledge node path data according to the nursing knowledge theme, which effectively organizes and simplifies the complex health care knowledge system and provides a basis for subsequent analysis and push. By loading the second knowledge node path data into the graph neural network model and generating the first graph coding feature, the complex relationship and structural information between the knowledge nodes can be captured and represented, which greatly improves the depth and breadth of data processing and analysis, and can more comprehensively understand the internal logic and association of health care knowledge. Subsequently, by calculating the first feature distance between the personalized nursing knowledge point data and the first graph coding feature of each second knowledge node path data, the key knowledge node path data that best matches the user's needs is identified in multiple second knowledge node path data, ensuring that the pushed nursing suggestions are not only personalized, but also highly relevant and effective. Finally, for the follow-up care stage corresponding to the key knowledge node path data, the nursing execution suggestion information matching the personalized nursing knowledge point data is pushed, thereby directly targeting the user's current health status and nursing needs, providing personalized nursing guidance for the user, thereby effectively improving the effect and quality of the user's health management. In this way, the personalized nursing suggestions are pushed accurately and efficiently. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be extracted in combination with these drawings without creative work.

[0014] Figure 1 A flow chart of a method for pushing page information for user health management provided in an embodiment of the present application;

[0015] Figure 2 This is a schematic block diagram of the architecture of the smart nursing system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0016] The following description is intended to enable one of ordinary skill in the art to implement and combine the present application, and the description is provided in the context of a specific application scenario and its requirements. It will be apparent to one of ordinary skill in the art that various changes may be made to the disclosed embodiments, and that the general principles defined in the present application may be applied to other embodiments and application scenarios without departing from the principles and scope of the present application. Therefore, the present application is not limited to the described embodiments, but should be given the broadest scope consistent with the claims.

[0017] Figure 1 This is a flow chart of a method for pushing page information applied to user health management provided by an embodiment of the present application. The method for pushing page information applied to user health management is introduced in detail below.

[0018] Step S110, integrating the knowledge node data corresponding to the same care migration stage in the user care health knowledge graph data and outputting them as first knowledge node path data, wherein the user care health knowledge graph data is a data set to be analyzed and personalized care recommendations to be pushed.

[0019] In detail, the user care health knowledge graph data is a set of structured data sets used to describe various knowledge and information related to user health care, which can be represented in the form of a graph, where nodes represent knowledge units (such as diseases, symptoms, care methods, etc.) and edges represent the relationship between these nodes. This graph data structure is suitable for representing complex health care knowledge systems. In the graph data structure, a knowledge node is an entity that represents specific health care knowledge. For example, a knowledge node may represent a disease, a drug, a care method, etc.

[0020] The nursing migration stage refers to the different stages of the patient's nursing process, such as the initial diagnosis stage, treatment stage, rehabilitation stage, etc. Each stage may require different nursing knowledge and methods.

[0021] The first knowledge node path data is a data path formed by integrating all knowledge node data corresponding to the same nursing migration stage, representing the concatenation of all relevant knowledge in a specific nursing stage.

[0022] In this embodiment, taking the health monitoring and care scenarios of the elderly and obese groups as an example, the user care health knowledge graph data contains a large amount of information related to the health care of these two groups of people. For example, for the elderly, it may involve knowledge nodes in cardiovascular disease prevention, osteoporosis care, cognitive function maintenance, etc.; for the obese group, it may include knowledge nodes such as obesity-induced diabetes prevention, joint stress management, and metabolic regulation. These knowledge node data are also associated with different stages of care migration, such as various stages from initial diagnosis to rehabilitation.

[0023] Taking cardiovascular disease prevention in the elderly as an example, in the preliminary assessment stage, knowledge nodes may include data related to basic health check-up items such as blood pressure measurement and blood lipid testing, as well as lifestyle surveys such as salt and fat intake in the diet, daily activity, etc. In the prevention and intervention stage, knowledge nodes include data on exercise recommendations (such as low-intensity aerobic exercise suitable for the elderly, such as walking, Tai Chi, etc., frequency and intensity requirements), dietary adjustments (specific guidance on reducing saturated fatty acid intake, increasing vegetable and fruit intake, etc.), and drug use (if there is a slight blood pressure problem, the type of drug, dosage and time of taking, etc.). For the obese population in diabetes prevention, the knowledge nodes in the preliminary assessment stage include body mass index (BMI) measurement, blood sugar testing, family medical history surveys, etc.; in the intervention stage, there are weight loss diet plans (such as specific plans to control carbohydrate intake), exercise plans (such as moderate-intensity exercise suitable for obese people, such as swimming and cycling plans), and other data.

[0024] The server will integrate these knowledge node data corresponding to the same care migration stage. For example, all knowledge node data of the elderly and obese groups in the preliminary assessment stage are integrated and output as the first knowledge node path data. This first knowledge node path data can be understood as a clue that runs through all relevant knowledge in the preliminary assessment stage, covering information from basic health checks to lifestyle surveys. Whether it is for the elderly or obese groups, as long as the knowledge nodes related to this stage are integrated into this path data.

[0025] Step S120, integrating the first knowledge node path data corresponding to the same nursing knowledge topic in the user nursing health knowledge graph data and outputting them as second knowledge node path data.

[0026] In detail, the nursing knowledge theme refers to a broad field or classification of health care knowledge, such as disease prevention, rehabilitation care, nutritional advice, etc. These themes help organize and classify health care knowledge. The second knowledge node path data is a data path formed by further integrating the first knowledge node path data corresponding to the same nursing knowledge theme, representing the integration of all relevant knowledge under a certain nursing knowledge theme.

[0027] That is, in this embodiment, the server will find the part corresponding to the same nursing knowledge topic from the first knowledge node path data obtained previously.

[0028] Taking the nursing knowledge theme of disease prevention as an example, the server will integrate and output the first knowledge node path data related to cardiovascular disease prevention for the elderly and diabetes prevention for obese groups as the second knowledge node path data. In this process, the server will analyze the various elements in the first knowledge node path data to determine which ones are related to the theme of disease prevention. For cardiovascular disease prevention in the elderly, blood pressure and blood lipid testing are to detect risks early and prevent disease onset, and exercise and diet adjustments are also to reduce the risk of cardiovascular disease; for diabetes prevention in obese groups, BMI measurement and blood sugar testing are to identify risks early, and weight loss diet plans and exercise plans are also key measures to prevent diabetes. These data closely related to the theme of disease prevention will be integrated into the second knowledge node path data. This second knowledge node path data can be understood as a comprehensive information collection that focuses more on the theme of disease prevention, covering various related knowledge of the elderly and obese groups under this theme.

[0029] Step S130: load the second knowledge node path data into a graph neural network model to generate a first graph encoding feature for expressing the second knowledge node path data.

[0030] In detail, the graph neural network model is a special neural network model specifically used to process graph structure data, which can learn the relationship between nodes and edges, so as to encode and represent the graph structure data. The first graph encoding feature is the encoding feature generated by the graph neural network model after processing the second knowledge node path data, which can express the complex relationship and structural information in the second knowledge node path data.

[0031] In this embodiment, the server loads the second knowledge node path data into the graph neural network model. For example, for the cardiovascular disease prevention part of the elderly in the second knowledge node path data, the blood pressure measurement data contained therein may be related to the exercise recommendation data (for example, the exercise intensity of the elderly with higher blood pressure needs to be adjusted more carefully), and the diet adjustment data is also related to the blood lipid test data (a high-fat diet may cause dyslipidemia, which in turn affects cardiovascular health, so dietary adjustments should be based on blood lipid test results). For the prevention of diabetes in obese groups, BMI data is related to weight loss diet plans (higher BMIs may require stricter dietary control), and exercise plans are also related to blood sugar data (proper exercise helps control blood sugar levels).

[0032] When processing the second knowledge node path data, the graph neural network model will take into account the complex relationship between these knowledge nodes. At the same time, the server also needs to obtain the collaborative annotation information corresponding to the second knowledge node path data from the user's care health knowledge graph data. In this scenario, taking the elderly as an example, the nursing operation information is extracted from the second knowledge node path data through the entity recognition network and the relationship extraction network. For example, identifying a nursing operation such as "walking for 30 minutes every day" involves the identification of entities in the relevant text (such as walking-nursing operation, 30 minutes every day-operation conditions) and the extraction of relationships (the time association between walking and 30 minutes). Then, through the user portrait analysis network, the user nursing portrait information corresponding to the nursing operation information is extracted from the second knowledge node path data, which may include age, physical function status (such as whether there are joint problems that affect walking) and other information for the elderly. Aggregate the nursing operation information and the user nursing portrait information to generate collaborative annotation information.

[0033] Then the second knowledge node path data and collaborative annotation information are loaded into the graph neural network model at the same time. For example, the second knowledge node path data containing disease prevention for the elderly and obese groups and the corresponding collaborative annotation information (such as the association annotations between the age, physical condition and cardiovascular disease prevention measures of the elderly, and the association annotations between the BMI, body metabolic condition and diabetes prevention measures of the obese group) are loaded into the model. The model generates a first graph encoding feature for expressing the second knowledge node path data by learning the relationship structure between these data. This first graph encoding feature can comprehensively represent the various relationships and features in the second knowledge node path data, and the health care knowledge related to both the elderly and the obese group is effectively expressed in this feature.

[0034] Step S140, obtain personalized nursing knowledge point data, determine the first feature distance between the personalized nursing knowledge point data and the first graph encoding feature of each second knowledge node path data, and determine the key knowledge node path data in multiple second knowledge node path data based on the first feature distance.

[0035] In detail, the personalized nursing knowledge point data is personalized health care information for a specific patient or user group, and can be customized based on the patient's personal health status, living habits, genetic information, etc. The first feature distance is an indicator of the degree of similarity or difference between the personalized nursing knowledge point data and the first graph encoding features of each second knowledge node path data. The smaller the first feature distance, the higher the similarity.

[0036] The key knowledge node path data (which is the data path that best matches the user's personalized nursing knowledge point data among the plurality of second knowledge node path data determined according to the first characteristic distance) represents the health care knowledge path that best suits the current patient.

[0037] In this embodiment, personalized care knowledge point data may be generated based on the special circumstances of a specific elderly person or obese individual. For example, for an elderly person with mild arthritis, personalized care knowledge point data may include adjustments to cardiovascular disease prevention exercise plans for arthritis (such as reducing exercise programs that put greater pressure on joints); for an obese individual with a family history of diabetes, personalized care knowledge point data may include stricter blood sugar monitoring frequency and special dietary adjustment recommendations.

[0038] The server then determines the first feature distance between the personalized nursing knowledge point data and the first graph coding feature of each second knowledge node path data. Taking the elderly as an example, multiple first nursing logic deduction knowledge are extracted from the personalized nursing knowledge point data. Assuming that the personalized nursing knowledge point data is about the adjustment of cardiovascular disease prevention exercise plans for arthritis in the elderly, the first nursing logic deduction knowledge therein may include "because arthritis affects joint movement, reduce long-term standing sports, such as some movements in Tai Chi". For this first nursing logic deduction knowledge, the server determines its first knowledge logic density in the first graph coding feature.

[0039] First, the first graph coding feature is decomposed into multiple semantic units. For example, the graph coding feature related to cardiovascular disease prevention is decomposed into semantic units such as blood pressure control, exercise type, and dietary structure, and its semantic type, location information, and semantic boundary information are marked. For the above-mentioned first nursing logic deduction knowledge, it is decomposed into components such as "arthritis affects joint activity" and "reduce long-term standing sports", and the semantic role of each component (the former is the logical starting point, and the latter is the logical result part), the logical connection method (causal relationship), etc., are determined, and the association relationship with the global semantic structure is established. By comparing the similarity of semantic connotation, semantic type, and logical role, a semantic mapping relationship with the semantic unit in the first graph coding feature is constructed, and the target area (such as the semantic unit area related to the exercise type) that is logically associated with the first nursing logic deduction knowledge is identified in the first graph coding feature. Then, the pre-constructed logic density calculation framework is called, and the evaluation value of the target area on each indicator is calculated according to the calculation rules of the evaluation indicators (such as the number of semantic units, the richness of logical relationships, and the degree of semantic matching). The preliminary logic density value is obtained by weighted processing, and then the final first knowledge logic density is obtained according to the relationship between the indicators.

[0040] Then determine the first knowledge logic weight of the first nursing logic derivation knowledge in multiple first knowledge representations. Perform semantic feature analysis on multiple first knowledge representations. For the first knowledge representation of cardiovascular disease prevention, its semantic features include information on blood pressure, blood lipids, exercise, diet, etc., and record its distribution parameters. For the above first nursing logic derivation knowledge, decompose its semantic content into "arthritis" (core semantic element), "joint activity impact" (logical relationship part), etc., and determine its semantic connotation, semantic boundary and semantic role. For each first knowledge representation, evaluate the semantic association with the first nursing logic derivation knowledge. For example, in the first knowledge representation of cardiovascular disease prevention, compare the semantic features of "arthritis" with blood pressure, blood lipids, etc., find the target semantic features related to "arthritis" semantics (such as the association between exercise type and joints), determine the semantic association score, combine the matching degree of logical relationship type to obtain the semantic association, and finally normalize it to obtain the first knowledge logic weight.

[0041] According to the first matching degree between each first nursing logic derivation knowledge and the first knowledge representation, the first characteristic distance between the personalized nursing knowledge point data and the first knowledge representation is determined. In this way, the key knowledge node path data is determined in the plurality of second knowledge node path data.

[0042] Step S150, pushing nursing execution suggestion information matching the personalized nursing knowledge point data for the subsequent nursing stage corresponding to the key knowledge node path data.

[0043] In detail, the nursing execution suggestion information is specific nursing advice or guidance provided to users based on key knowledge node path data, aiming to help users or medical staff implement effective health care measures.

[0044] That is, after determining the key knowledge node path data, the server pushes nursing execution suggestion information matching the personalized nursing knowledge point data for the subsequent nursing stage corresponding to the key knowledge node path data.

[0045] Taking the obese group as an example, if the key knowledge node path data is about diabetes prevention for obese individuals, and the individual has a family history of diabetes and has experienced slight blood sugar fluctuations, the personalized care knowledge point data includes stricter blood sugar monitoring frequency and special diet adjustment suggestions. If the current stage is the intervention stage, the follow-up care stage may be to further strengthen the intervention or transition to the rehabilitation stage.

[0046] The nursing execution suggestion information pushed by the server may include more detailed dietary adjustment plans, such as specific food types to increase dietary fiber intake (such as eating more oats, beans, etc.), and more precise exercise time and intensity adjustments (such as adding 15 minutes of moderate-intensity exercise time per day). The same is true for the elderly. If the key knowledge node path data is about the prevention of cardiovascular disease in elderly people with mild arthritis, and the follow-up care stage is the continuous prevention period, the nursing execution suggestion information pushed by the server may be based on arthritis. Further optimization of exercise methods (such as increasing upper limb exercise to reduce lower limb joint pressure) and adjustment of dietary structure (such as increasing foods rich in calcium and vitamin D to protect joints and maintain cardiovascular health) and so on. In this way, more accurate and personalized health monitoring and care suggestions can be provided to the elderly and obese groups.

[0047] Based on the above steps, the embodiment of the present application realizes the accurate push of personalized nursing suggestions by integrating the relevant information in the user's nursing health knowledge graph data. Specifically, the knowledge node data is first integrated into the first knowledge node path data according to the nursing migration stage, and then the first knowledge node path data is further integrated into the second knowledge node path data according to the nursing knowledge theme, which effectively organizes and simplifies the complex health care knowledge system and provides a basis for subsequent analysis and push. By loading the second knowledge node path data into the graph neural network model and generating the first graph coding feature, the complex relationship and structural information between the knowledge nodes can be captured and represented, which greatly improves the depth and breadth of data processing and analysis, and can more comprehensively understand the internal logic and association of health care knowledge. Subsequently, by calculating the first feature distance between the personalized nursing knowledge point data and the first graph coding feature of each second knowledge node path data, the key knowledge node path data that best matches the user's needs is identified in multiple second knowledge node path data, ensuring that the pushed nursing suggestions are not only personalized, but also highly relevant and effective. Finally, for the follow-up care stage corresponding to the key knowledge node path data, the nursing execution suggestion information matching the personalized nursing knowledge point data is pushed, thereby directly targeting the user's current health status and nursing needs, providing personalized nursing guidance for the user, thereby effectively improving the effect and quality of the user's health management. In this way, the personalized nursing suggestions are pushed accurately and efficiently.

[0048] In a possible implementation, step S130 includes:

[0049] Step S131, obtaining collaborative annotation information corresponding to the second knowledge node path data from the user care health knowledge graph data.

[0050] Step S132: Load the second knowledge node path data and the collaborative annotation information into the graph neural network model simultaneously to generate the first graph encoding feature for expressing the second knowledge node path data.

[0051] In a possible implementation, step S131 includes:

[0052] Step S1311, extracting nursing operation information from the second knowledge node path data through an entity recognition network and a relationship extraction network.

[0053] Step S1312: extracting user nursing portrait information corresponding to the nursing operation information from the second knowledge node path data through a user portrait analysis network.

[0054] Step S1313, aggregating the nursing operation information and the user nursing portrait information corresponding to the nursing operation information to generate the collaborative annotation information.

[0055] In this embodiment, the server can use the entity recognition network and the relationship extraction network to extract nursing operation information from the second knowledge node path data. Taking the prevention of cardiovascular diseases in the elderly as an example, the second knowledge node path data contains relevant content such as blood pressure detection, exercise plan, and diet adjustment. Through the entity recognition network, it is possible to identify entity information such as "measurement" in "measure blood pressure once a day" as a nursing operation, "blood pressure" as an operation object, and "once a day" as an operation frequency. The relationship extraction network will further clarify the relationship between these entities, such as the association between blood pressure test results and exercise plan adjustments. If the blood pressure is high, the exercise intensity may need to be reduced. For the prevention of diabetes in obese groups, in the diet plan in the second knowledge node path data, the entity recognition network can identify "control" in "control carbohydrate intake" as an operation, and "carbohydrate intake" as an operation object. The relationship extraction network can determine the relationship between the diet plan and weight control, blood sugar stability and other nursing operation information.

[0056] At the same time, the server can also extract user care profile information corresponding to the care operation information from the second knowledge node path data through the user portrait analysis network. For the elderly, user care profile information may include age, gender, whether there are other chronic diseases (such as whether arthritis affects exercise ability), family medical history, etc. For example, if an elderly person has a family history of hypertension, then the care operations for cardiovascular disease prevention (such as the choice of exercise intensity, control of salt in the diet, etc.) need to be formulated more carefully. For obese groups, user care profile information may include height, weight, body metabolic rate, whether there are other complications (such as whether there are joint problems that affect exercise) and family history of diabetes. For example, if an obese individual has a family history of diabetes and a low body metabolic rate, then special considerations must be given to the care operations for diabetes prevention (such as the strictness of dietary adjustments, the choice of exercise types, etc.).

[0057] The server then aggregates the extracted nursing operation information and the corresponding user nursing portrait information to generate collaborative annotation information. For the prevention of cardiovascular diseases in the elderly, the collaborative annotation information may be "70-year-old male, with a family history of hypertension, blood pressure test once a day, moderate exercise intensity (considering age and family history), and strict control of salt intake in the diet (based on family history)". For the prevention of diabetes in obese groups, the collaborative annotation information may be "weight 90 kg, height 170 cm, low body metabolic rate and family history of diabetes, carbohydrate intake needs to be strictly controlled (considering family history and body metabolism), and low-impact exercise types (based on weight and possible joint problems)".

[0058] Afterwards, the server loads the second knowledge node path data and the collaborative annotation information into the graph neural network model at the same time, thereby generating a first graph encoding feature for expressing the second knowledge node path data. In the scenario of cardiovascular disease prevention in the elderly, the second knowledge node path data contains blood pressure detection, exercise plan, diet adjustment and other knowledge and the relationship between them, and the collaborative annotation information contains the association between the individual characteristics of the elderly and the nursing operations. The graph neural network model will learn the intrinsic structure and relationship of these data. For example, the graph neural network model will learn that with age (obtained from the collaborative annotation information), blood pressure fluctuations (obtained from the blood pressure detection-related knowledge in the second knowledge node path data) may be more frequent, so that more flexible adjustments are needed in exercise plans (exercise plan-related knowledge in the second knowledge node path data), and salt control in the diet (diet adjustment-related knowledge in the second knowledge node path data) also needs to be stricter. These relationships and features will be encoded in the first graph encoding feature.

[0059] For diabetes prevention in obese groups, the knowledge of weight, diet plan, exercise plan, etc. in the second knowledge node path data is loaded into the graph neural network model together with the information of individual height, weight, family medical history, etc. in the collaborative annotation information. The model will analyze that due to the low body metabolic rate (collaborative annotation information), in terms of weight control (weight-related knowledge in the second knowledge node path data), the carbohydrate control in the diet plan (diet plan-related knowledge in the second knowledge node path data) should be more precise, and the intensity and type of exercise in the exercise plan (exercise plan-related knowledge in the second knowledge node path data) should also be adjusted according to weight and possible joint problems (collaborative annotation information). These relationships are finally encoded into the first graph encoding features to comprehensively express the various characteristics and relationships of the second knowledge node path data in the health monitoring and care scenarios of the elderly and obese groups.

[0060] In a possible implementation, step S132 includes:

[0061] Step S1321, identify various structural elements in the second knowledge node path data through the entity recognition network, and after formatting and converting the various structural elements, perform preliminary division of the formatted conversion data into semantic units based on the professional vocabulary library and semantic segmentation rules in the nursing field to generate different semantic units. The semantic unit is a single word, phrase or sentence structure, and the various structural elements include different levels of labels, indexes and existing metadata information.

[0062] In this embodiment, taking the second knowledge node path data related to the prevention of cardiovascular diseases in the elderly as an example, these structural elements include different levels of labels, indexes, and existing metadata information. For example, the blood pressure detection record in the second knowledge node path data may have a time tag (such as the blood pressure value measured at 8 o'clock in the morning). This time tag is a structural element, as well as index information about the blood pressure measurement device (for example, using a specific model of sphygmomanometer) and metadata information attached to the blood pressure value itself (such as physical state information during measurement, etc.). For the second knowledge node path data related to the prevention of diabetes in obese groups, there may be structural elements such as the time tag of the weight measurement record, the index of the weight scale, and metadata information such as body posture during weight measurement. After the server formats and converts these various structural elements, it performs preliminary division of the formatted conversion data into semantic units based on the professional vocabulary library and semantic segmentation rules in the nursing field. In the scenario of cardiovascular disease prevention for the elderly, "the blood pressure value measured by an Omron blood pressure monitor at 8 a.m. was 130 / 80 mmHg, and the body was in a calm state during the measurement" may be divided into different semantic units such as "8 a.m." (single lexical semantic unit), "Omron blood pressure monitor" (phrase semantic unit), "measured blood pressure value" (phrase semantic unit), "130 / 80 mmHg" (single lexical semantic unit), "the body was in a calm state" (sentence structure semantic unit). For diabetes prevention for obese people, "the weight measured by a body fat scale at 3 p.m. was 90 kg, and the feet were balanced during the measurement" may be divided into semantic units such as "3 p.m.", "body fat scale", "measured weight", "90 kg", and "balanced feet".

[0063] Step S1322, pre-construct a corpus, and use the corpus to pre-annotate each semantic unit in the second knowledge node path data, the corpus contains text materials related to pre-nursing operations, and the pre-annotation process is achieved by calculating the similarity between each semantic unit and different semantic categories in the corpus, wherein for each semantic unit, after converting the semantic unit into a feature vector representation through a word vector model, calculate the similarity between the feature vector of the semantic unit and the different semantic category vectors pre-marked in the implementation corpus, and pre-annotate each semantic unit as a corresponding semantic category according to the similarity.

[0064] Next, the server pre-builds a corpus that contains text materials related to pre-care operations, and then uses the corpus to pre-label each semantic unit in the second knowledge node path data. For semantic units in cardiovascular disease prevention for the elderly, such as the semantic unit "measuring blood pressure value", the server converts it into a representation of a feature vector through a word vector model. Assume that the word vector model converts this semantic unit into a multidimensional vector, and then calculates the similarity between this vector and the different semantic category vectors pre-labeled in the corpus. If there is a semantic category "health detection operation" in the corpus, and the feature vector of "measuring blood pressure value" has a high similarity with the semantic category vector of "health detection operation", then "measuring blood pressure value" is pre-labeled as the semantic category of "health detection operation". For the semantic unit "measuring weight" in diabetes prevention for obese people, it is also converted into a feature vector through a word vector model and compared with the semantic category vector in the corpus for similarity, and may be pre-labeled as the semantic category of "body index measurement operation".

[0065] Step S1323, for each semantic unit with a pre-labeled semantic category, analyze the context environment of the semantic unit, perceive the context environment of the semantic unit through a pre-built context semantic model to calculate the dynamic correlation of the semantic unit with different semantic categories in the current context, and based on the dynamic correlation of the semantic unit with different semantic categories in the current context, the context environment includes the previous and next semantic units of the semantic unit and the semantic categories of the previous and next semantic units, and the context semantic model is implemented based on a neural network or a semantic analysis model based on a graph structure.

[0066] Afterwards, for each semantic unit with a pre-labeled semantic category, the server perceives the context environment of the semantic unit through the pre-built context semantic model to calculate the dynamic correlation of the semantic unit with different semantic categories in the current context. In the scenario of cardiovascular disease prevention for the elderly, for the semantic unit "measuring blood pressure value", if the semantic units before and after it are "8 o'clock in the morning" and "130 / 80mmHg", and the semantic categories of the semantic units before and after are "time" and "blood pressure value", the context semantic model (implemented by a neural network or a graph-based semantic analysis model) will analyze this context environment. Assuming that in this case, the model finds that the time "8 o'clock in the morning" is related to "measuring blood pressure value" in a specific nursing operation logic (for example, blood pressure may be relatively high in the morning), then the dynamic correlation of "measuring blood pressure value" with the semantic category "blood pressure measurement time-related operation" in the current context will increase. For the semantic unit "measuring weight" in the prevention of diabetes in obese groups, if the previous and next semantic units are "3 pm" and "90 kg", the contextual semantic model will analyze the impact of this contextual environment on the semantic category, and may find that the weight in the afternoon will fluctuate relative to the morning, thereby affecting the dynamic correlation of the semantic unit "measuring weight" in different semantic categories. According to the dynamic correlation of the semantic unit with different semantic categories in the current context, the server adjusts the pre-annotated semantic category of the semantic unit and generates the second knowledge node path data of the re-annotated semantic category. For example, after context analysis, "measuring blood pressure value" may be adjusted from the initially pre-annotated "health check operation" to a more precise semantic category of "morning blood pressure measurement operation"; "measuring weight" may be adjusted from "body index measurement operation" to "afternoon weight measurement operation".

[0067] Step S1324, according to the dynamic correlation between the semantic unit and different semantic categories in the current context, the pre-annotated semantic category of the semantic unit is adjusted, and after the second knowledge node path data of the re-annotated semantic category is generated, the potential entities in the second knowledge node path data of the re-annotated semantic category are identified using potential entity recognition rules, wherein the potential entities refer to various objects related to the nursing operation, including the executor of the nursing operation, the object of the nursing operation, the tools used for the nursing operation, and the relevant environmental elements of the nursing operation.

[0068] Then, the server uses the latent entity recognition rules to identify the latent entities in the second knowledge node path data that are annotated with semantic categories. In the prevention of cardiovascular diseases in the elderly, for the data "the blood pressure value measured in the morning using an Omron blood pressure monitor is 130 / 80 mmHg", the latent entity recognition rules will identify "Omron blood pressure monitor" as the tool used for nursing operations (the tool entity of nursing operations), "blood pressure value" as the object of nursing operations (the object entity of nursing operations), and the person who performs the blood pressure measurement operation may be a medical staff or the elderly themselves (the executor entity of nursing operations). "Morning" and "the body is in a calm state" during measurement can be regarded as relevant environmental factors of nursing operations (the environmental entity of nursing operations). In the prevention of diabetes in obese groups, for "the weight is measured in the afternoon using a body fat scale and is 90 kg", "body fat scale" will be identified as a tool entity, "weight" as an object entity, the executor may also be a medical staff or the obese individual himself, and "afternoon" and "standing on both feet in balance" as environmental entities.

[0069] Step S1325, after identifying the potential entities, based on the semantic similarity of the potential entities and the relevance of the potential entities in the nursing operation, the potential entities are preliminarily clustered to generate at least one entity cluster.

[0070] After identifying the potential entities, the server performs preliminary clustering of the potential entities based on the semantic similarity of each potential entity and the relevance of each potential entity in the nursing operation to generate at least one entity cluster. In the scenario of cardiovascular disease prevention for the elderly, for potential entities related to blood pressure measurement, "Omron blood pressure monitor" and possible other blood pressure measurement devices (if there are multiple device records) will be clustered into one entity cluster based on the semantic similarity of them being measurement tools and their relevance in blood pressure measurement operations; "blood pressure value" and possible other blood pressure-related indicators (such as diastolic pressure and systolic pressure analyzed separately, etc.) will be clustered into one entity cluster about the measurement object; the executor (medical staff or the elderly themselves) may form an entity cluster separately; environmental factors such as "morning" and "the body is in a calm state" may be clustered into one environmental entity cluster. For diabetes prevention in obese groups, measurement tools such as "body fat scales" are clustered into one entity cluster, physical indicators such as "weight" are clustered into one entity cluster, executors are clustered into one entity cluster, and environmental factors such as "afternoon" and "standing on both feet" are clustered into one entity cluster.

[0071] Step S1326, through the relationship extraction network, analyze the grammatical structure and semantic logical relationship in the second knowledge node path data of the re-annotated semantic category, and identify the relationship pattern between the entities in the at least one entity cluster based on the grammatical structure and semantic logical relationship, the relationship pattern includes a subject-predicate relationship pattern, a verb-object relationship pattern or a modifier-predicate relationship pattern.

[0072] Subsequently, the server uses the relationship extraction network to analyze the grammatical structure and semantic logical relationship in the second knowledge node path data of the semantic category, and identifies the relationship pattern between the entities in at least one entity cluster based on the grammatical structure and semantic logical relationship. In the prevention of cardiovascular diseases in the elderly, for the entity cluster containing "Omron blood pressure monitor" and "blood pressure value", according to the grammatical structure "Omron blood pressure monitor measures blood pressure value", it can be identified that this is a subject-object relationship pattern; for the entity cluster containing the executor and "measure blood pressure value", it may be a subject-predicate relationship pattern (such as medical staff measuring blood pressure value). In the prevention of diabetes in obese groups, for the entity clusters of "body fat scale" and "weight", the subject-object relationship pattern is identified based on the grammatical structure of "body fat scale measures weight"; the relationship between the executor and "measure weight" may be a subject-predicate relationship pattern.

[0073] Step S1327: for each of the entity clusters, corresponding relationship candidate information is constructed based on the identified relationship pattern, where the relationship candidate information is an expression describing the relationship between entities.

[0074] For each entity cluster, the server constructs corresponding relationship candidate information based on the identified relationship pattern. These relationship candidate information are expressions that describe the relationship between entities. In the prevention of cardiovascular diseases in the elderly, for the entity cluster "Omron blood pressure monitor measures blood pressure value" with a subject-object relationship pattern, the relationship candidate information may be "blood pressure monitor-measurement-blood pressure value"; for the actor and the "measurement blood pressure value" entity cluster with a subject-predicate relationship pattern, the relationship candidate information may be "actor-act-measurement blood pressure value". In the prevention of diabetes in obese groups, for the entity cluster "body fat scale measures weight" with a subject-object relationship pattern, the relationship candidate information is "body fat scale-measurement-weight"; for the actor and the "measurement weight" entity cluster with a subject-predicate relationship pattern, the relationship candidate information is "actor-act-measurement weight".

[0075] Step S1328, pre-establish a nursing operation logic knowledge base, the nursing operation logic knowledge base contains nursing operation logic rules in the nursing field, for each combination data of the entity cluster and relationship candidate information, a verification and adjustment process is performed according to the nursing operation logic knowledge base, and nursing operation information is extracted from the adjusted combination data, the nursing operation information includes the specific name of the nursing operation, the executor of the operation, the object of the operation, the tool used for the operation and the relevant conditions of the operation.

[0076] Finally, for each entity cluster and relationship candidate information combination data, the verification and adjustment process is carried out according to the nursing operation logic knowledge base, and the nursing operation information is extracted from the adjusted combination data. In the prevention of cardiovascular diseases in the elderly, for the combination data of "Omron blood pressure monitor - measurement - blood pressure value", the nursing operation logic knowledge base may stipulate the accuracy range of different types of blood pressure monitors for measuring blood pressure and other rules. If the measurement result of the Omron blood pressure monitor is within a reasonable accuracy range, then this combination data is valid. From this valid combination data, it can be extracted that the specific name of the nursing operation is "blood pressure measurement", the executor of the operation (as mentioned above, it may be medical staff or the elderly themselves), the object of the operation is "blood pressure value", the tool used for the operation is "Omron blood pressure monitor" and the relevant conditions of the operation (such as the body is in a calm state during measurement, etc.). In the prevention of diabetes in obese groups, for the combined data of "body fat scale - measurement - weight", the rules about the measurement accuracy and measurement posture of body fat scale in the nursing operation logic knowledge base are verified and adjusted. If it is effective, the specific name of the nursing operation can be extracted as "weight measurement", the executor (medical staff or obese individuals themselves), the object of the operation is "weight", the tool used for the operation is "body fat scale" and the relevant conditions of the operation (such as standing balance on both feet, etc.). Through such a series of complex and delicate steps, the server successfully extracted the nursing operation information from the second knowledge node path data. Whether it is for the prevention of cardiovascular diseases in the elderly or the prevention of diabetes in obese groups, it can accurately obtain various key information related to nursing operations, providing important data support for subsequent health monitoring and nursing work.

[0077] In a possible implementation, step S132 may further include:

[0078] Step A110, fuses the collaborative annotation information into the first guidance information to generate first fused annotation information, wherein the first guidance information includes metadata, configuration information or indicative data related to the processing of the graph neural network model, which is used to enhance the graph neural network model's ability to understand and encode the knowledge node path data.

[0079] In this embodiment, taking the prevention of cardiovascular diseases in the elderly as an example, the collaborative annotation information may include the elderly's age (such as 75 years old), gender (male), whether there are other chronic diseases (such as mild arthritis), family history (such as family history of hypertension), etc. The first guidance information contains metadata, configuration information or indicative data related to the processing of the graph neural network model, such as the parameter settings of the graph neural network model for health care data processing (such as weight allocation parameters for different data types), data format requirements, etc. Fusion of collaborative annotation information into the first guidance information, for the prevention of cardiovascular diseases in the elderly, may be to combine the information such as the age and gender of the elderly with the parameter settings of the model processing. For example, if the model has different processing weights for cardiovascular disease prevention data of different age groups, the age information of 75 years old will be associated with the corresponding weight parameters; if the model treats male and female data differently in certain feature analysis, the gender information will also be fused with the configuration information of such differential treatment. Family history and other chronic disease information will be fused with indicative data (such as special processing instructions for family history or chronic diseases) to generate the first fused annotation information. For the prevention of diabetes in obese groups, the weight (such as 90 kg), height (such as 170 cm), body metabolic rate (such as low), whether there is a family history of diabetes (if any) and other contents in the collaborative annotation information are fused with the first guiding information (including specific parameters and format requirements of the graph neural network model for obesity-related health care data processing). For example, weight and height information may be fused with relevant parameters in the model for body mass index (BMI) calculation and analysis, and body metabolic rate and family history information may be fused with indicative data for special health risk processing for obese individuals, thereby generating the first fused annotation information for the prevention of diabetes in obese groups.

[0080] Step A120: Load the second knowledge node path data and the first fusion annotation information into the graph neural network model simultaneously to generate the first graph encoding feature for expressing the second knowledge node path data.

[0081] In a possible implementation, step A120 includes:

[0082] Step A121: extract multiple first focused knowledge node data from the second knowledge node path data, and generate a first focused knowledge node vector based on the first focused knowledge node data.

[0083] Step A122, edit multiple of the first focused knowledge node vectors into a first input feature representation for loading into the graph neural network model.

[0084] Step A123, based on the first fused annotation information, generate a second input feature representation for loading into the graph neural network model.

[0085] Step A124, loading the first input feature representation and the second input feature representation into the graph neural network model simultaneously, generating the first graph encoding feature for expressing the second knowledge node path data.

[0086] Next, the server loads the second knowledge node path data and the first fusion annotation information into the graph neural network model simultaneously to generate the first graph encoding feature. This process includes multiple sub-steps.

[0087] First, the server extracts multiple first focused knowledge node data from the second knowledge node path data, and generates a first focused knowledge node vector based on these first focused knowledge node data. In the scenario of cardiovascular disease prevention in the elderly, the second knowledge node path data contains knowledge node data in various aspects such as blood pressure detection, blood lipid detection, exercise plan, and diet adjustment. The server may identify key data in blood pressure detection (such as normal range of blood pressure, blood pressure fluctuation, etc.), core elements in exercise plan (such as exercise type, exercise frequency, exercise intensity, etc.) as first focused knowledge node data. For the normal range in blood pressure detection (such as systolic blood pressure 90-139mmHg, diastolic blood pressure 60-89mmHg), it is converted into a form that can be used for calculation and representation, such as being expressed in vector form as [90, 139, 60, 89] (here is just a simple illustration of the vector representation form), and the exercise type (such as walking, Tai Chi) can be converted into a vector by a pre-set encoding method, such as walking is encoded as [1, 0], Tai Chi is encoded as [0, 1], etc., and then the first focused knowledge node vector is generated based on these first focused knowledge node data. For diabetes prevention in obese groups, the weight control target, blood sugar test results, carbohydrate intake in the diet plan, exercise intensity and time in the exercise plan, etc. in the second knowledge node path data are identified as the first focused knowledge node data. The weight control target (such as losing 5 kg) can be converted into a vector form, the blood sugar test results (such as fasting blood sugar value) are also converted into the corresponding vector, the carbohydrate intake (such as 150 grams per day), exercise intensity (such as the metabolic equivalent value corresponding to moderate intensity exercise) and time (such as 30 minutes of exercise per day) are all converted into vectors through a specific encoding method, and then these vectors are combined to generate the first focused knowledge node vector.

[0088] Afterwards, the server edits multiple first focused knowledge node vectors into a first input feature representation for loading into the graph neural network model. In the prevention of cardiovascular diseases in the elderly, the first focused knowledge node vectors related to blood pressure detection, exercise plan, etc. generated previously are edited according to a certain order and rules. For example, if the corresponding vectors are arranged and combined in the order of blood pressure detection, exercise plan, blood lipid detection, etc., and some separation marks may be added or normalized to form a complete first input feature representation, this first input feature representation can comprehensively cover the information of key knowledge nodes related to the prevention of cardiovascular diseases in the elderly, and is presented in a format suitable for the input of the graph neural network model. For the prevention of diabetes in obese groups, the first focused knowledge node vectors related to weight control goals, blood sugar test results, diet plans and exercise plans are edited in a specific order (such as the order of weight-blood sugar-diet-exercise), and the first input feature representation for loading into the graph neural network model is generated by adding necessary connectors or performing data standardization operations, so that this representation can accurately convey the core knowledge node information related to the prevention of diabetes in obese groups to the graph neural network model.

[0089] Then, the server generates a second input feature representation for loading into the graph neural network model based on the first fused annotation information. For the first fused annotation information for the prevention of cardiovascular diseases in the elderly (including age, gender, family medical history and other information fused with the graph neural network model processing), this information is encoded and converted in a way that the graph neural network model can understand. For example, age 75 can be encoded as a specific value according to the age range set by the model, gender male can be encoded as 0 or 1 (assuming 0 represents female and 1 represents male), and family medical history (with a history of hypertension) can be encoded as a vector according to the type of family medical history and the degree of impact on cardiovascular disease. These encoded information are combined according to a certain structure to form a second input feature representation. For the prevention of diabetes in obese groups, information such as weight, height, body metabolic rate, family history of diabetes, etc. in the first fused annotation information is also similarly encoded and converted. A weight of 90 kg and a height of 170 cm can be calculated as a BMI value and encoded as a specific numerical value, the body metabolic rate (low) is encoded as a value according to the pre-set metabolic rate level, and the family history of diabetes (yes) is encoded as a vector according to its impact on the risk of diabetes. These encoded data are then combined into the second input feature representation.

[0090] Finally, the server loads the first input feature representation and the second input feature representation into the graph neural network model at the same time, and generates a first graph encoding feature for expressing the second knowledge node path data. In the scenario of cardiovascular disease prevention in the elderly, the first input feature representation containing core knowledge node information such as blood pressure detection and exercise plan and the second input feature representation containing fusion information such as age, gender, and family history are simultaneously input into the graph neural network model. The graph neural network model will learn the relationship between these input features based on its own structure and algorithm. For example, the model will learn the relationship between age 75 (obtained from the second input feature representation) and blood pressure fluctuations (obtained from the blood pressure detection information in the first input feature representation), the relationship between family history (history of hypertension) and exercise plan (exercise plan information in the first input feature representation), etc. By learning and processing these relationships, a first graph encoding feature is generated that can fully express the second knowledge node path data related to cardiovascular disease prevention in the elderly. This first graph encoding feature contains the complex relationship between knowledge nodes in various aspects of cardiovascular disease prevention and the association information with individual characteristics of the elderly. For diabetes prevention in obese groups, the first input feature representation containing information such as weight control goals, blood sugar test results, diet plans and exercise plans and the second input feature representation containing information such as weight, height, body metabolic rate, family diabetes history, etc. are loaded into the graph neural network model at the same time. The model analyzes the relationship between weight (obtained from the second input feature representation) and weight control goals (obtained from the first input feature representation), the relationship between family diabetes history and blood sugar test results, diet plans and exercise plans, etc., so as to generate the first graph encoding feature that can accurately express the second knowledge node path data related to diabetes prevention in obese groups. This feature integrates the characteristics of obese individuals and various relationship information between knowledge nodes related to diabetes prevention. In this way, the server successfully generated the first graph encoding feature for the second knowledge node path data related to cardiovascular disease prevention in the elderly and diabetes prevention in obese groups, providing an important data basis for subsequent health monitoring and care analysis.

[0091] In a possible implementation, the first graph encoding feature is a first knowledge representation of the second knowledge node path data.

[0092] Step S140 includes:

[0093] Step S141, extracting a plurality of first nursing logic derivation knowledge from the personalized nursing knowledge point data.

[0094] In this embodiment, taking the prevention of cardiovascular disease in the elderly as an example, it is assumed that the personalized nursing knowledge point data contains the detailed health status and nursing needs of a specific elderly person. If this elderly person not only has a risk of cardiovascular disease, but also suffers from mild arthritis, then the personalized nursing knowledge point data may contain information on the adjustment of the cardiovascular disease prevention nursing plan based on the arthritis condition. The first nursing logic derivation knowledge extracted from this data may include: because arthritis affects joint activity, in the exercise plan for cardiovascular disease prevention, long-term standing or excessive knee bending should be avoided; and because arthritis may affect the mobility and metabolism of the elderly, dietary adjustments should focus more on nutritional balance to support joint repair and cardiovascular health. For the prevention of diabetes in obese groups, assuming that the personalized nursing knowledge point data is for an obese individual with a specific family history, the first nursing logic derivation knowledge may include: in view of the family history of diabetes and the current obesity status, the frequency of blood glucose monitoring should be higher than that of ordinary obese individuals; and because the individual metabolic rate is low, in the weight loss diet plan, carbohydrate intake should be strictly controlled while increasing protein intake to maintain basal metabolism.

[0095] Step S142: for each piece of the first nursing logic-derived knowledge, determine a first matching degree between the first nursing logic-derived knowledge and the first knowledge representation.

[0096] Next, for the first nursing logic derivation knowledge in cardiovascular disease prevention for the elderly, such as the aforementioned exercise plan adjustment under the influence of arthritis. The server will perform a matching analysis on this knowledge with the first knowledge representation (the representation of the second knowledge node path data related to cardiovascular disease prevention for the elderly generated by the graph neural network model). The first knowledge representation contains information such as the relationship between blood pressure, blood lipids, exercise and diet, cardiovascular disease risk assessment and nursing measures for different age groups, etc. The server will analyze in detail the logical derivation knowledge of exercise plan adjustment and the exercise-related part in the first knowledge representation. For example, in the first knowledge representation, if the normal cardiovascular disease prevention exercise recommendation is to take a certain length of walk and a moderate amount of joint activity exercises every day, and the knowledge derived from the personalized nursing knowledge point data is to avoid certain movements due to arthritis, the server will compare the logical associations and differences between the two. Calculate the logical proportion of this arthritis-based exercise plan adjustment knowledge in the entire first knowledge representation, the logical coherence with other related knowledge (such as the relationship between blood pressure control and exercise), and other factors to determine the first matching degree of the first nursing logic derivation knowledge in the first knowledge representation. For the first nursing logic derivation knowledge in diabetes prevention for obese groups, such as blood sugar monitoring frequency adjustment and diet plan adjustment under special circumstances, the server also performs a matching analysis with the first knowledge representation (representation of the second knowledge node path data related to diabetes prevention for obese groups). The first knowledge representation contains information such as the relationship between blood sugar and weight, diet, and exercise, and diabetes prevention strategies under different obesity levels and family histories. The server will analyze the relationship between the blood sugar monitoring frequency adjustment knowledge and the general recommendations for blood sugar monitoring in the first knowledge representation, as well as the matching degree between the diet plan adjustment knowledge and the relationship between the diet structure, nutrients, and diabetes prevention in the first knowledge representation, and determine the first matching degree from multiple aspects such as logical structure and knowledge association.

[0097] Step S143: determining the first characteristic distance between the personalized nursing knowledge point data and the first knowledge representation according to the first matching degree between each of the first nursing logic-derived knowledge and the first knowledge representation.

[0098] Finally, in the scenario of cardiovascular disease prevention in the elderly, if the first nursing logic derivation knowledge about the impact of arthritis on exercise plan adjustment and the first knowledge representation have a low first matching degree, while the first nursing logic derivation knowledge about dietary adjustment to support joint and cardiovascular health and the first knowledge representation have a high first matching degree, the server will comprehensively consider these different matching conditions. The first characteristic distance between the personalized nursing knowledge point data and the first knowledge representation is calculated through a specific algorithm (such as the weight distribution and distance calculation rules according to the matching degree). This first characteristic distance reflects the degree of difference between the representation of the personalized nursing knowledge point data and the second knowledge node path data related to cardiovascular disease prevention in the elderly generated by the graph neural network model. The same is true for diabetes prevention in obese groups. Taking into account the different matching degrees of the first nursing logic derivation knowledge such as blood sugar monitoring frequency adjustment and diet plan adjustment and the first knowledge representation, the first characteristic distance between the personalized nursing knowledge point data and the first knowledge representation is determined according to the established algorithm, thereby providing important data basis for subsequent personalized health monitoring and nursing decisions for the elderly and obese groups.

[0099] In a possible implementation, step S141 includes:

[0100] Step S1411, pre-constructing a semantic analysis framework, wherein the semantic analysis framework is constructed based on the professional knowledge system in the nursing field, and includes definitions and descriptions of various concepts, entities, and relationship semantic elements in the nursing field.

[0101] In this embodiment, this semantic analysis framework is constructed based on the professional knowledge system in the nursing field, covering the definitions and descriptions of many concepts, entities, and relational semantic elements related to health care for the elderly and obese groups. For example, in the prevention of cardiovascular diseases in the elderly, concepts may include blood pressure, blood lipids, atherosclerosis, etc.; entities may include elderly individuals, medical staff, nursing equipment (such as sphygmomanometers), etc.; relational semantic elements include the relationship between blood pressure and cardiovascular disease risk, the impact of blood lipid levels on vascular health, etc. For the prevention of diabetes in obese groups, concepts include body mass index (BMI), blood sugar, insulin resistance, etc., entities include obese individuals, nutritionists, fitness coaches, etc., and relational semantic elements include the relationship between BMI and the risk of diabetes, and the association between blood sugar control and diet and exercise.

[0102] Step S1412, dividing the personalized nursing knowledge point data into semantic areas according to the semantic analysis framework, and generating data of different semantic areas or semantic relationship parts.

[0103] Taking cardiovascular disease prevention in the elderly as an example, if the personalized care knowledge point data contains information such as the physical condition of an elderly person (such as mild arthritis, family history of hypertension), living habits (such as dietary preferences, daily activity), and recent health check results (such as blood pressure, blood lipid values). Then according to the semantic analysis framework, the physical condition can be divided into a semantic area, which can be further divided into sub-semantic relationship parts such as disease conditions (arthritis, family history); living habits are divided into another semantic area, including two sub-semantic relationship parts of diet and activity; health check results are a separate semantic area. For diabetes prevention in obese groups, if the personalized care knowledge point data involves the weight, height, family history of diabetes, daily diet structure, exercise frequency, and blood sugar monitoring data of an obese individual, the semantic area division may be that physical indicators (weight, height, blood sugar) are one area, family history is a separate area, and lifestyle (diet structure, exercise frequency) is one area, and each area contains its own related semantic relationship parts.

[0104] Step S1413, decompose the data in each semantic area or semantic relationship part into multiple semantic units, analyze the semantic connotation and semantic boundary of each semantic unit, determine the semantic characteristics of the semantic unit in the nursing semantic system, and analyze the internal relationship between each semantic unit based on the semantic characteristics of each semantic unit in the nursing semantic system, and establish a mapping relationship between each semantic unit.

[0105] For the blood pressure values ​​in the semantic area of ​​health check results for cardiovascular disease prevention in the elderly, this data can be decomposed into semantic units such as systolic blood pressure values ​​and diastolic blood pressure values. For the semantic unit of systolic blood pressure values, its semantic connotation is analyzed to indicate the pressure value generated by blood on the blood vessel wall when the heart contracts, and the semantic boundary is the normal blood pressure range (such as 90-139mmHg) that defines the reasonable range of this value. Determine the semantic characteristics of this semantic unit in the nursing semantic system, for example, it is one of the key indicators reflecting the cardiovascular function state. Then, based on the semantic characteristics of each semantic unit in the nursing semantic system, analyze the internal relationship between each semantic unit and establish a mapping relationship. For example, there is a correlation between systolic blood pressure values ​​and diastolic blood pressure values. They jointly reflect the blood pressure state and have a logical connection with the risk of cardiovascular disease, thereby establishing a mapping relationship between them. For the semantic area of ​​dietary structure in diabetes prevention for obese people, data such as food types and nutrients are decomposed into semantic units. For example, the carbohydrate content in a certain food is a semantic unit. Its semantic connotation is analyzed to indicate the amount of carbohydrates in the food, and the semantic boundary is defined by the reasonable range of daily carbohydrate intake of the human body. Its semantic characteristics are important factors affecting blood sugar levels. By analyzing the relationship with other nutrient semantic units (such as protein and fat content), a mapping relationship between them is established. For example, carbohydrates and proteins have a mutual relationship in energy supply and metabolism.

[0106] Step S1414, based on the mapping relationship between each semantic unit, search and identify the actual logical relationship instances in the personalized nursing knowledge point data, for each logical relationship instance, determine the semantic units involved in the logical relationship instance, and construct the logical relationship type between each semantic unit in the logical relationship instance, wherein in the process of constructing the logical relationship type, the logical relationship across semantic areas is integrated.

[0107] In terms of cardiovascular disease prevention in the elderly, taking the two semantic areas of physical condition and health check results as examples, the logical relationship instance may be: due to the family history of hypertension (semantic unit in the semantic area of ​​physical condition), the systolic blood pressure value (semantic unit in the semantic area of ​​health check results) is in a higher risk range. For this logical relationship instance, the semantic units involved are determined to be the family history of hypertension and the systolic blood pressure value, and the logical relationship type between them is constructed as a causal relationship. In the process of constructing the logical relationship type, if a logical relationship across semantic areas is involved, such as the relationship between dietary preferences in living habits (such as high-salt diet) and blood pressure values ​​in health check results, it is integrated and determined to be a causal relationship in which diet affects blood pressure. For diabetes prevention in obese groups, for example, there is a logical relationship instance between the BMI value (semantic unit) in physical indicators and the family history of diabetes (semantic unit) in family medical history. Due to a family history of diabetes, the risk of diabetes increases when the BMI is high (obesity). The logical relationship type between the two semantic units is determined to be a causal relationship, and the logical relationship instance between the dietary structure (such as high-sugar diet) and the blood sugar value is integrated to determine the causal relationship in which diet affects blood sugar.

[0108] Step S1415, pre-constructing a set of candidate logical deduction knowledge patterns, wherein each logical deduction knowledge pattern in the set of candidate logical deduction knowledge patterns contains corresponding logical elements and logical relationship types between the logical elements.

[0109] Later, in the scenario of cardiovascular disease prevention for the elderly, the logical deduction knowledge patterns in this set may include: family medical history (logical element) - influence - disease risk (logical element), living habits (logical element) - change - health indicators (logical element) and other patterns, each of which contains corresponding logical elements and the types of logical relationships between them. For diabetes prevention in obese people, there may be logical deduction knowledge patterns such as family medical history - increase - disease risk, dietary structure - influence - blood sugar level.

[0110] Step S1416, matches and associates the logical relationship type between each of the semantic units in the logical relationship instance with the set of candidate logical deduction knowledge patterns, and for each logical relationship instance, searches for a matching target logical deduction knowledge pattern in the set of candidate logical deduction knowledge patterns, wherein the matching process is achieved by comparing the logical element type, logical relationship type and global logical structure in the logical relationship instance with the association between the corresponding parts in the logical deduction knowledge pattern.

[0111] For the causal relationship logical relationship instance between family history of hypertension and systolic blood pressure value in cardiovascular disease prevention for the elderly, a matching target logical deduction knowledge pattern is searched in the candidate logical deduction knowledge pattern set. By comparing the logical element type (family history and health indicators), logical relationship type (causal relationship), and global logical structure (family history affects health indicators) in the logical relationship instance, it is associated with the corresponding part in the logical deduction knowledge pattern. It is found that the logical deduction knowledge pattern of family history-influence-disease risk matches it. For the causal relationship logical relationship instance between BMI value and family history of diabetes in diabetes prevention for obese people, the logical element type (physical indicators and family history), logical relationship type (causal relationship), and global logical structure are also compared to find the matching target logical deduction knowledge pattern of family history-increase-risk of disease.

[0112] Step S1417, extracting nursing logic deduction knowledge from the personalized nursing knowledge point data according to the target logic deduction knowledge pattern matched by each logical relationship instance, wherein, for each logical relationship instance marked with the target logic deduction knowledge pattern, the corresponding nursing logic deduction knowledge is extracted from the logical relationship instance according to the logical element definition in the target logic deduction knowledge pattern.

[0113] Finally, for the logical relationship instance (family history of hypertension and systolic blood pressure value) of the target logical deduction knowledge model marked with family history-influence-disease risk in cardiovascular disease prevention for the elderly, the corresponding nursing logical deduction knowledge is extracted from the logical relationship instance according to the logical element definition in this model. For example, because there is a family history of hypertension, it is necessary to pay closer attention to blood pressure values ​​in nursing and take more active preventive measures (such as stricter diet control, more regular exercise, etc.) to reduce the risk of cardiovascular disease. For the logical relationship instance (BMI value and family history of diabetes) of the target logical deduction knowledge model marked with family history-increase-risk of disease in diabetes prevention for obese people, the extracted nursing logical deduction knowledge is: because there is a family history of diabetes and a high BMI, it is necessary to strengthen blood sugar monitoring, adjust the diet structure (reduce carbohydrate intake), and increase exercise to reduce the risk of diabetes in nursing. Through such a series of steps, the server can accurately extract multiple first nursing logical deduction knowledge from personalized nursing knowledge point data, providing an important basis for subsequent health monitoring and nursing decisions.

[0114] In a possible implementation, step S142 includes:

[0115] Step S1421: for each of the first nursing logic-derived knowledge, determine the first knowledge logic density of the first nursing logic-derived knowledge in the first knowledge representation.

[0116] In this embodiment, taking the prevention of cardiovascular disease in the elderly as an example, it is assumed that the first nursing logic derivation knowledge is to adjust the exercise plan for cardiovascular disease prevention based on the arthritis condition of the elderly, that is, to reduce the sports items that put greater pressure on the joints. The first knowledge representation contains various information on the prevention of cardiovascular disease in the elderly, such as the relationship between blood pressure, blood lipids and exercise diet, cardiovascular disease risk assessment and nursing measures for different age groups, etc. The server decomposes the first knowledge representation into multiple semantic units, such as blood pressure values, blood lipid indicators, exercise types, exercise frequency, dietary ingredients, etc., and marks the semantic type of each semantic unit (such as blood pressure values ​​as physiological indicator types, exercise types as nursing operation types), position information in the first knowledge representation (such as blood pressure values ​​in the health indicator part, exercise types in the nursing measures part) and semantic boundary information (such as the normal range of blood pressure values, classification boundaries of exercise types). For the first nursing logic derivation knowledge of the above-mentioned exercise plan adjustment, it is decomposed into components such as "arthritis affects joint activity" and "reduce sports items with greater joint pressure", and the semantic role of each component (the former is the logical starting point, and the latter is the logical result part), the logical connection method (causal relationship), etc., is determined, and an association relationship with the global semantic structure is established. By comparing the similarity of semantic connotation, semantic type and logical role, a semantic mapping relationship between the semantic units in the first knowledge representation is constructed. In the first knowledge representation, a target area that is logically associated with the first nursing logic deduction knowledge is identified, such as a semantic unit area related to the exercise type. Then, a pre-constructed logic density calculation framework is retrieved. This framework contains evaluation indicators of multiple dimensions, calculation rules and weight distribution rules for each evaluation indicator, and the relationship between each evaluation indicator. The evaluation indicators include the number of semantic units, the richness of the logical relationship between semantic units, and the semantic matching degree between the logical association area and the first nursing logic deduction knowledge as a whole. According to the calculation rules of these evaluation indicators, the evaluation value of the target area on each indicator is calculated respectively. For example, in the semantic unit area related to the exercise type, the number of semantic units may be a limited number of exercise types, the richness of the logical relationship may be reflected in the relationship between different exercise types and cardiovascular health, and the semantic matching degree is to see the degree of matching between these exercise types and the logical result of reducing joint stress. The evaluation values ​​of each evaluation indicator are weighted to obtain a preliminary logical density value, and the preliminary logical density value is adjusted according to the relationship between each evaluation indicator to obtain the final first knowledge logical density. For diabetes prevention in obese people, it is assumed that the first nursing logic derivation knowledge is to adjust the carbohydrate intake in the diet plan according to the lower body metabolic rate of obese individuals. The first knowledge representation includes information such as the relationship between blood sugar and body weight, diet, and exercise, and diabetes prevention strategies under different obesity levels and family history.The above operations are also performed to decompose the first knowledge representation and the first nursing logic derivation knowledge, construct a semantic mapping relationship, identify the target area (such as the semantic unit area related to the diet plan), and calculate the first knowledge logic density through the logic density calculation framework.

[0117] Step S1422: determine the first knowledge logic weight of the first nursing logic-derived knowledge in the plurality of first knowledge representations.

[0118] In the scenario of cardiovascular disease prevention in the elderly, there are multiple first knowledge representations, which may be cardiovascular disease prevention knowledge representations for the elderly of different genders and different health conditions (such as whether they have other chronic diseases). The server performs semantic feature analysis on these first knowledge representations, obtains the semantic features in each first knowledge representation, and records the distribution parameters of each semantic feature in each first knowledge representation. For example, for the first knowledge representation for male elderly people without other chronic diseases, its semantic features may include the normal range of physiological indicators such as blood pressure and blood lipids, suitable exercise types (such as slightly higher intensity exercise), etc. The distribution parameters are the proportion of these semantic features in this knowledge representation (such as 30% of blood pressure and blood lipid-related content, 40% of exercise type-related content, etc.) and the logical relationship type between semantic features and other semantic features (such as the mutual influence relationship between exercise type and blood pressure and blood lipids). For the first nursing logic derivation knowledge based on arthritis adjustment exercise plan, its semantic content is decomposed into "arthritis" (core semantic element), "joint activity influence" (logical relationship part), etc., to determine its semantic connotation, semantic boundary and semantic role in nursing logic derivation knowledge. For each first knowledge representation, the semantic association between the first nursing logic derivation knowledge and the first knowledge representation is evaluated. For example, in the first knowledge representation for male elderly people without other chronic diseases, the semantic features of "arthritis" are compared with those of blood pressure, blood lipids, exercise type, etc., and the target semantic features semantically related to "arthritis" (such as the association between exercise type and joints) are found to determine the semantic association score. Combined with the matching degree of the logical relationship type between each core semantic element and the logical relationship type between each semantic feature in the first knowledge representation, the semantic association score is weighted and calculated to obtain the semantic association between the first nursing logic derivation knowledge and each first knowledge representation. Finally, the obtained semantic association is normalized to obtain the first knowledge logic weight of the first nursing logic derivation knowledge in the first knowledge representation. For the prevention of diabetes in obese groups, the above operations are also performed for the first knowledge representations with different obesity levels and different family history combinations, and the semantic features are analyzed, the semantic association is calculated, and the first knowledge logic weight is obtained by normalization.

[0119] Step S1423: determining the first matching degree according to the first knowledge logic density and the first knowledge logic weight.

[0120] Finally, in the prevention of cardiovascular diseases in the elderly, the first knowledge logic density and the first knowledge logic weight of the first nursing logic deduction knowledge based on the arthritis adjustment exercise plan calculated above are comprehensively considered. If the first knowledge logic density is high, it means that the logic deduction knowledge in the first knowledge representation is closely logically related and rich in content; if the first knowledge logic weight is also high, it means that the logic deduction knowledge in multiple first knowledge representations has a strong correlation. Through a specific calculation method (such as weighted average, etc.), the two are combined to determine the first matching degree. This first matching degree can accurately reflect the matching degree between the first nursing logic deduction knowledge and the first knowledge representation, and provide an important basis for subsequent health monitoring and nursing decisions. For the prevention of diabetes in obese groups, the first matching degree between the first nursing logic deduction knowledge and the first knowledge representation is also determined based on the calculated first knowledge logic density and first knowledge logic weight.

[0121] In one possible implementation, step S1421, comprising;

[0122] Step S1421-1, decompose the first knowledge representation into multiple semantic units, and mark the semantic type, position information in the first knowledge representation, and semantic boundary information of each semantic unit.

[0123] Step S1421-2, decompose the first nursing logic derivation knowledge into multiple components, where different components are different semantic concepts, logical relationship elements or knowledge fragments.

[0124] Step S1421-3, determine the semantic role of each of the components, the logical connection method between the components, and the semantic boundaries and semantic connotations between the components, and establish an association relationship between each component and the global semantic structure of the first nursing logic deduction knowledge. The semantic role includes a logical starting point, a logical intermediate link, and a logical result part. The logical connection method includes a causal relationship, a progressive relationship, and a parallel relationship.

[0125] Step S1421-4, by comparing the similarities of semantic connotations, semantic types and logical roles, construct a semantic mapping relationship between the various components of the first nursing logic deduction knowledge and the semantic units in the first knowledge representation, and based on the semantic mapping relationship, identify a target area that is logically associated with the first nursing logic deduction knowledge in the first knowledge representation, wherein for each semantic unit in the first knowledge representation in the mapping relationship, a target area is identified with the semantic unit as the center according to the logical connection relationship between the semantic units, and the target area contains other semantic units that are logically related to the semantic unit.

[0126] Step S1421-5, call up a pre-built logic density calculation framework, which includes evaluation indicators of multiple dimensions, calculation rules and weight allocation rules for each evaluation indicator, and the relationship between each evaluation indicator. The evaluation indicators include the number of semantic units, the richness of the logical relationship between semantic units, and the semantic matching degree between the logical association area and the first nursing logic deduction knowledge as a whole.

[0127] Step S1421-6, based on the logic density calculation framework, calculate the logic density of the first nursing logic deduction knowledge in the first knowledge representation, wherein the evaluation value of the target area on each evaluation indicator is calculated separately according to the calculation rules of each evaluation indicator, and the evaluation value of each evaluation indicator is weighted according to the weight distribution principle to obtain a preliminary logic density value, and the preliminary logic density value is adjusted according to the relationship between each evaluation indicator to obtain the final first knowledge logic density.

[0128] In this embodiment, in terms of cardiovascular disease prevention in the elderly, it is assumed that the first knowledge representation contains comprehensive information ranging from blood pressure and blood lipid monitoring, different cardiovascular disease risk factors to various preventive care measures (such as exercise, dietary recommendations, etc.).

[0129] The server first decomposes the first knowledge representation into multiple semantic units. For example, a blood pressure value (such as 130 / 80mmHg) is a semantic unit, and its semantic type is a physiological indicator. Its position information in the first knowledge representation may be in the health status assessment part, and its semantic boundary information is the normal blood pressure range (90 - 139 / 60 - 89mmHg). Walking in the exercise type, the semantic type is the exercise category in the nursing measures, the position information is in the preventive nursing measures part, and the semantic boundary is the definition of walking as a mode of exercise (definition of speed, rhythm, etc.). Salt intake in the diet, the semantic type is the dietary composition in the nursing measures, the position information is in the dietary recommendations part, and the semantic boundary is the reasonable range of daily salt intake (such as no more than 5 grams), etc.

[0130] Assume that the first nursing logic derivation knowledge is to adjust the exercise plan for cardiovascular disease prevention based on the joint problems of the elderly. Decompose it into multiple components, such as "joint pain affects mobility" (this is a semantic concept, as a logical starting point), "reduce high-impact exercise" (this is a logical relationship element, indicating action adjustment, belonging to the logical result part), "select low-intensity joint-friendly exercise" (this is a knowledge fragment, also a logical result part). Determine the semantic role of each component. "Joint pain affects mobility" is the logical starting point, which is the reason for the subsequent adjustment of the exercise plan; "reduce high-impact exercise" and "select low-intensity joint-friendly exercise" are the logical result parts, which are the response measures generated based on the logical starting point. The logical connection method is causal relationship. The exercise plan needs to be adjusted because of joint pain. At the same time, determine the semantic boundaries and semantic connotations between each component. The semantic connotation of "joint pain affects mobility" is the limitation of joint pain on the daily activity ability of the elderly. The semantic boundary is the degree range of joint pain (such as mild and moderate pain) and the specific manifestations of its impact on mobility (such as shortened walking distance, difficulty going up and down stairs, etc.).

[0131] Establish the association between each component and the global semantic structure of the first nursing logic deduction knowledge. "Joint pain affects mobility" is the starting point of the entire logic deduction knowledge, which determines the necessity of subsequent exercise plan adjustment; "reducing high-impact exercise" and "selecting low-intensity joint-friendly exercise" are specific response measures based on the former, together forming the complete logic of adjusting exercise plans for joint problems in the elderly.

[0132] By comparing the similarity of semantic connotation, semantic type and logical role, the semantic mapping relationship between each component of the first nursing logic deduction knowledge and the semantic unit in the first knowledge representation is constructed. For the component "joint pain affects mobility", semantic units with similar semantic connotations are searched in the first knowledge representation, and semantic units related to physical function conditions may be found, such as joint flexibility assessment results. Because they all involve the description of the relevant conditions of the body joints, there is a certain correlation in semantic types (both are related to physical conditions), and the joint flexibility assessment results may be used as a reference factor for judging the impact of joint pain on mobility in terms of logical roles. For "reducing high-impact exercise" and "selecting low-intensity joint-friendly exercise", there is a semantic mapping relationship with semantic units related to exercise types (such as running, Tai Chi and other exercise types and their characteristics) in the first knowledge representation, because these semantic units involve the type and intensity of exercise and are directly related to adjusting the exercise plan.

[0133] Based on these semantic mapping relationships, the target area that is logically associated with the first nursing logic deduction knowledge is identified in the first knowledge representation. Taking the semantic unit of Tai Chi in the exercise type as an example, a target area is identified based on the logical connection relationship between semantic units. This area may contain semantic units such as the stress assessment of joints by Tai Chi, the relationship between Tai Chi and cardiovascular health, and the physical condition suitable for practicing Tai Chi. Because these semantic units are logically related to the exercise type of Tai Chi, they together constitute the target area that is logically associated with the first nursing logic deduction knowledge.

[0134] The pre-built logic density calculation framework is retrieved. This logic density calculation framework contains evaluation indicators of multiple dimensions, such as the number of semantic units, the richness of the logical relationship between semantic units, the semantic matching degree between the logical association area and the first nursing logic deduction knowledge as a whole, etc., and stipulates the calculation rules and weight distribution rules of each evaluation indicator and the relationship between each evaluation indicator.

[0135] Specifically, according to the calculation rules of each evaluation indicator, the evaluation value of the target area on each indicator is calculated separately. For the number of semantic units, in the target area centered on Tai Chi, the number of related semantic units is counted, assuming 10. In terms of the richness of logical relationships between semantic units, there are many logical relationships between Tai Chi and joint pressure, cardiovascular health, and physical condition. For example, Tai Chi has little joint pressure, which is beneficial to the elderly with joint pain (causal relationship), and moderate Tai Chi practice is conducive to cardiovascular health (causal relationship). According to the complexity and diversity of these relationships, an evaluation value is given, assuming 0.6 (a value between 0-1). In terms of the semantic matching degree between the logical association area and the first nursing logical deduction knowledge as a whole, because the semantic units in the target area are highly correlated with the logical deduction knowledge of adjusting the exercise plan based on joint problems (such as joint pressure, cardiovascular health, etc. are related to the adjustment of the exercise plan), a higher evaluation value is given, assuming 0.8.

[0136] According to the weight distribution principle, the evaluation values ​​of each evaluation index are weighted to obtain the preliminary logical density value. Assuming that the weight of the number of semantic units is 0.3, the weight of the richness of logical relations is 0.3, and the weight of the semantic matching degree is 0.4, the preliminary logical density value = 10 0.3+0.6 0.3 + 0.8 0.4 = 3 + 0.18+0.32 = 3.5. Then, the preliminary logic density value is adjusted according to the relationship between the various evaluation indicators. Assuming that the semantic matching degree has a greater impact on the overall logic density, according to the pre-set adjustment rules (such as when the semantic matching degree is greater than 0.7, the preliminary logic density value is increased by a certain proportion), the final first knowledge logic density is 4.0.

[0137] In terms of diabetes prevention in obese groups, it is assumed that the first knowledge representation includes information ranging from body weight, body metabolic rate, blood sugar monitoring to diet and exercise intervention.

[0138] For example, a weight value (such as 90 kg) is a semantic unit, with a semantic type of physiological indicator, location information in the physical condition assessment section, and a semantic boundary that defines the weight range for different degrees of obesity. The semantic type of body metabolic rate (such as low metabolic rate) is physiological function indicator, location information in the physical function assessment section, and semantic boundary is the standard for dividing normal metabolic rate and low metabolic rate. The semantic type of dietary carbohydrate intake (such as 200 grams per day) is dietary component, location information in the diet plan section, and semantic boundary is the reasonable carbohydrate intake range determined based on weight and metabolic rate, etc.

[0139] Assume that the first nursing logic derivation knowledge is to adjust the diet plan for diabetes prevention based on the low metabolic rate of obese individuals. Decompose it into "low metabolic rate affects energy consumption" (logical starting point, semantic concept), "reduce carbohydrate intake" (logical result, logical relationship element), and "increase protein intake to improve metabolism" (logical result, knowledge fragment). Determine the semantic role of each component. "Low metabolic rate affects energy consumption" is the logical starting point and is the reason for adjusting the diet plan; "reduce carbohydrate intake" and "increase protein intake to improve metabolism" are the logical result parts and are measures to deal with low metabolic rate. The logical connection method is causal relationship. Determine the semantic boundaries and semantic connotations between each component. The semantic connotation of "low metabolic rate affects energy consumption" is that low metabolic rate leads to a reduced ability of the body to consume energy. The semantic boundary is the specific numerical range of low metabolic rate and the quantitative expression of its impact on energy consumption (such as consuming 100 kcal less per day, etc.).

[0140] Establish the association relationship between each component and the global semantic structure of the first nursing logic deduction knowledge. "Low metabolic rate affects energy consumption" determines the need to adjust the diet plan, and "reducing carbohydrate intake" and "increasing protein intake to improve metabolism" are specific diet adjustment strategies for controlling blood sugar and improving metabolism based on low metabolic rate.

[0141] By comparing the similarity of semantic connotation, semantic type and logical role, the semantic mapping relationship between each component of the first nursing logic deduction knowledge and the semantic unit in the first knowledge representation is constructed. For "low metabolic rate affects energy consumption", there is a semantic mapping relationship between the semantic units related to the body's metabolic rate in the first knowledge representation (such as the measurement method of metabolic rate, factors affecting metabolic rate, etc.), because they all revolve around metabolic rate and have related semantic types. The measurement method and influencing factors of metabolic rate in logical roles can be used as a reference for judging the impact of low metabolic rate on energy consumption. For "reducing carbohydrate intake" and "increasing protein intake to improve metabolism", there is a semantic mapping relationship between the semantic units related to dietary components in the first knowledge representation (such as the type of carbohydrates, the source of protein, etc.), because these semantic units involve the adjustment of specific components in the diet plan.

[0142] Based on these semantic mapping relationships, a target area that is logically associated with the first nursing logic deduction knowledge is identified in the first knowledge representation. Taking the semantic unit of carbohydrate intake as an example, a target area is identified based on the logical connection relationship between semantic units. This area may contain semantic units such as the relationship between carbohydrate intake and blood sugar, the basis for adjusting carbohydrate intake according to the body's metabolic rate, and the effects of different types of carbohydrates on blood sugar. Because these semantic units are logically related to carbohydrate intake, they together constitute a target area that is logically associated with the first nursing logic deduction knowledge.

[0143] The pre-built logic density calculation framework is retrieved, and the evaluation value of the target area on each indicator is calculated according to the calculation rules of each evaluation indicator. For the number of semantic units, it is assumed that there are 8 related semantic units in the target area centered on carbohydrate intake. In terms of the richness of logical relationships between semantic units, there are many logical relationships between carbohydrate intake and blood sugar, body metabolic rate, and carbohydrate types, such as excessive carbohydrate intake will increase blood sugar (causal relationship), and carbohydrate intake needs to be adjusted when the metabolic rate is low (causal relationship), etc. The evaluation value is given according to the complexity and diversity of these relationships, assuming it is 0.5. In terms of the semantic matching degree between the logical association area and the first nursing logical deduction knowledge as a whole, because the semantic units in the target area are highly correlated with the logical deduction knowledge of adjusting the diet plan based on low metabolic rate (such as blood sugar and metabolic rate are related to diet adjustment), a higher evaluation value is given, assuming it is 0.7.

[0144] According to the weight distribution principle, the evaluation values ​​of each evaluation indicator are weighted to obtain the preliminary logical density value. Assuming that the weight of the number of semantic units is 0.3, the weight of the richness of logical relations is 0.3, and the weight of the semantic matching degree is 0.4, the preliminary logical density value = 8 0.3+0.5 0.3 + 0.7 0.4 = 2.4+0.15 + 0.28 = 2.83. Then, the preliminary logic density value is adjusted according to the relationship between the various evaluation indicators. Assuming that the semantic matching degree has a greater impact on the overall logic density, according to the pre-set adjustment rules (such as when the semantic matching degree is greater than 0.6, the preliminary logic density value is increased by a certain proportion), the final first knowledge logic density is 3.2.

[0145] In a possible implementation, step S1422 includes:

[0146] Step S1422-1, perform semantic feature analysis on multiple first knowledge representations, obtain semantic features in each first knowledge representation, and record distribution parameters of each semantic feature in each first knowledge representation, wherein the semantic features are determined by analyzing key semantic units in the first knowledge representation, logical relationships between the key semantic units, and the global semantic structure, and the distribution parameters include the proportion of each semantic feature in a first knowledge representation, and the type of logical relationship between the semantic feature and other semantic features.

[0147] Step S1422-2, decompose the semantic content of the first nursing logic deduction knowledge into core semantic elements and logical relationship parts, and determine the semantic connotation, semantic boundary and semantic role of each core semantic element in the nursing logic deduction knowledge, as well as the logical relationship type between each core semantic element.

[0148] Step S1422-3, for each first knowledge representation, evaluate the semantic association between the first nursing logic deduction knowledge and the first knowledge representation, wherein the core semantic elements of the first nursing logic deduction knowledge are specifically compared with the semantic features in the first knowledge representation. For each core semantic element, by comparing the semantic connotation, semantic type and semantic role of the core semantic element, find the target semantic feature semantically related to the core semantic element in the first knowledge representation, and determine the semantic association score corresponding to the core semantic element and the corresponding target semantic feature. Combined with the matching degree of the logical relationship type between each core semantic element and the logical relationship type between each semantic feature in the first knowledge representation, the semantic association score is weightedly calculated to obtain the semantic association between the first nursing logic deduction knowledge and each first knowledge representation.

[0149] Step S1422-4, normalize the obtained semantic association to obtain the first knowledge logic weight of the first nursing logic derivation knowledge in the first knowledge representation.

[0150] In this embodiment, the server performs semantic feature analysis on multiple first knowledge representations. Assume that there are two first knowledge representations, one is a knowledge representation of cardiovascular disease prevention for elderly men without other chronic diseases, and the other is a knowledge representation of cardiovascular disease prevention for elderly women with mild hypertension.

[0151] For the first knowledge representation of elderly men without other chronic diseases, its semantic features are determined by analyzing the key semantic units (such as normal blood pressure and blood lipid values, suitable types of exercise such as brisk walking, etc.), the logical relationship between key semantic units (such as brisk walking helps maintain normal blood pressure and blood lipids), and the global semantic structure. Semantic features include normal blood pressure and blood lipid ranges, characteristics of exercise types suitable for men, etc. The distribution parameters of each semantic feature in this first knowledge representation are recorded, such as the semantic content related to the normal blood pressure and blood lipid range accounts for 30%, the content related to the exercise type accounts for 40%, etc., as well as the logical relationship type between the semantic feature and other semantic features, such as the relationship between the exercise type and blood pressure and blood lipids that promotes each other to maintain normal levels.

[0152] For the first knowledge representation of elderly women with mild hypertension, the semantic features include the blood pressure range of mild hypertension, the types of exercise suitable for women and considering hypertension (such as slower walking), etc. The distribution parameters are that the content related to the blood pressure range accounts for 40%, the content related to the type of exercise accounts for 30%, etc., and the logical relationship type is that the type of exercise has a positive impact on controlling blood pressure, etc.

[0153] Then, the semantic content of the first nursing logic derivation knowledge (adjusting the exercise plan for cardiovascular disease prevention based on joint problems in the elderly) is decomposed into core semantic elements and logical relationship parts. The core semantic element is "joint problems", whose semantic connotation is that the elderly have joint pain, limited mobility and other problems, and the semantic boundary is the severity of joint problems (mild, moderate, etc.). The semantic role in the nursing logic derivation knowledge is the logical starting point, which is the key factor that triggers the adjustment of the exercise plan. The logical relationship part is "adjusting the exercise plan based on joint problems", which indicates the causal relationship type between the core semantic element and the exercise plan.

[0154] For each first knowledge representation, evaluate the semantic association between the first nursing logic derivation knowledge and the first knowledge representation. For the first knowledge representation of elderly men without other chronic diseases, compare the core semantic element "joint problems" of the first nursing logic derivation knowledge with the semantic features in the knowledge representation. In this knowledge representation, look for the target semantic features that are semantically related to "joint problems". It may be found that the semantic features related to the type of exercise are related to joint problems (such as different types of exercise have different pressures on joints). Determine the semantic association score between "joint problems" and the corresponding target semantic feature (type of exercise), assuming it is 0.3. Combine the matching degree of the logical relationship type between each core semantic element (adjusting the exercise plan based on joint problems) and the logical relationship type between each semantic feature in the first knowledge representation (the relationship between exercise type and blood pressure and blood lipids). Since the role of exercise type in the two relationships is different (one is for joint pressure and the other is for blood pressure and blood lipids), the matching degree is low, assuming it is 0.2. The semantic association score is weighted to obtain the semantic association between the first nursing logic derivation knowledge and the first knowledge representation. Assuming the weighting coefficients are 0.5 (semantic association score) and 0.5 (logical relationship matching degree), the semantic association = 0.3 0.5+0.2 0.5 = 0.15 + 0.1 = 0.25.

[0155] For the first knowledge representation of elderly women with mild hypertension, the semantic features of "joint problems" are also compared with the knowledge representation. The target semantic feature found to be semantically related to "joint problems" may be the type of exercise (because the choice of exercise type for hypertensive patients also needs to consider joint problems), and the semantic association score is assumed to be 0.4. Combined with the matching degree of the logical relationship type, since the choice of exercise type for hypertensive patients and the adjustment of exercise plans for joint problems have certain commonalities (both factors such as joint pressure must be considered), the matching degree is assumed to be 0.6. The semantic association degree is obtained by weighted calculation. Assuming the same weighting coefficient, the semantic association degree = 0.4 0.5+0.6 0.5 = 0.2+0.3 = 0.5.

[0156] The obtained semantic relevance is normalized to obtain the first knowledge logic weight of the first nursing logic derivation knowledge in the first knowledge representation. 0.25 and 0.5 are normalized, and the total weight is assumed to be 1. The first knowledge logic weight in the first knowledge representation of elderly men without other chronic diseases is calculated to be 0.25 / (0.25 + 0.5) = 0.33, and the first knowledge logic weight in the first knowledge representation of elderly women with mild hypertension is 0.5 / (0.25 + 0.5) = 0.67.

[0157] Furthermore, for diabetes prevention in obese groups, it is assumed that there are two first knowledge representations, one is the diabetes prevention knowledge representation for obese people with normal body metabolic rate, and the other is the diabetes prevention knowledge representation for obese people with a family history of diabetes.

[0158] For the first knowledge representation of obesity with normal body metabolism rate, its semantic features are determined by analyzing key semantic units (such as weight, blood sugar value, suitable exercise types such as moderate-intensity exercise, etc.), logical relationships between key semantic units (such as moderate-intensity exercise helps control blood sugar), and global semantic structure. Semantic features include weight management strategies under normal body metabolism rate, the relationship between exercise and blood sugar control, etc. Record distribution parameters, such as 30% of the content related to weight management strategy, 40% of the content related to the relationship between exercise and blood sugar control, etc., as well as the type of logical relationship between semantic features and other semantic features, such as the mutual influence relationship between weight management and exercise (exercise helps weight management).

[0159] For the first knowledge representation of obesity and family history of diabetes, the semantic features include the impact of family history of diabetes on the risk of disease, special blood glucose monitoring strategies for family history, etc. The distribution parameters are that family history-related content accounts for 40%, blood glucose monitoring strategy-related content accounts for 30%, etc. The logical relationship type is the frequency and strictness of blood glucose monitoring affected by family history, etc.

[0160] The semantic content of the first nursing logic derivation knowledge (adjusting the diet plan for diabetes prevention based on the low metabolic rate of obese individuals) is decomposed into core semantic elements and logical relationship parts. The core semantic element is "low metabolic rate", whose semantic connotation is the low ability of the body to metabolize energy, and the semantic boundary is the specific numerical range of low metabolic rate. The semantic role in the nursing logic derivation knowledge is the logical starting point, which is the reason for adjusting the diet plan. The logical relationship part is "adjusting the diet plan based on low metabolic rate", which indicates the causal relationship type between the core semantic element and the diet plan.

[0161] For each first knowledge representation, evaluate the semantic association between the first nursing logic deduction knowledge and the first knowledge representation. For the first knowledge representation of obesity and normal body metabolic rate, compare the core semantic element "low metabolic rate" of the first nursing logic deduction knowledge with the semantic features in the knowledge representation. Look for the target semantic features semantically related to "low metabolic rate" in this knowledge representation. It may be found that weight management strategy has a certain association with low metabolic rate (because metabolic rate affects weight management). Determine the semantic association score between "low metabolic rate" and the corresponding target semantic feature (weight management strategy), assuming it is 0.2. Combine the matching degree of the logical relationship type between each core semantic element (adjusting the diet plan based on low metabolic rate) and the logical relationship type between each semantic feature in the first knowledge representation (the relationship between weight management and exercise). Since the relationship between the two is not directly related, the matching degree is low, assuming it is 0.1. Perform weighted calculation on the semantic association score to obtain the semantic association between the first nursing logic deduction knowledge and this first knowledge representation. Assuming the weighting coefficients are 0.5 (semantic association score) and 0.5 (logical relationship matching degree), the semantic association = 0.2 0.5+0.1 0.5= 0.1+0.05 = 0.15.

[0162] For the first knowledge representation of obesity and family history of diabetes, the semantic features of "low metabolic rate" are also compared with the semantic features in this knowledge representation. The target semantic feature found to be semantically related to "low metabolic rate" may be the blood glucose monitoring strategy (because low metabolic rate may affect blood glucose levels and is related to blood glucose monitoring strategy), and the semantic association score is assumed to be 0.3. Combined with the matching degree of the logical relationship type, since there is a certain correlation between the low metabolic rate affecting blood glucose and the family history affecting blood glucose monitoring strategy (both are related to blood glucose), the matching degree is assumed to be 0.4. The semantic association degree is obtained by weighted calculation. Assuming the weighting coefficients are the same, the semantic association degree = 0.3 0.5+0.4 0.5 = 0.15+0.2 = 0.35.

[0163] The obtained semantic relevance is normalized to obtain the first knowledge logic weight of the first nursing logic derivation knowledge in the first knowledge representation. 0.15 and 0.35 are normalized, and the total weight is assumed to be 1. The first knowledge logic weight in the first knowledge representation of obesity and normal body metabolism is calculated to be 0.15 / (0.15 + 0.35) = 0.3, and the first knowledge logic weight in the first knowledge representation of obesity and family history of diabetes is 0.35 / (0.15 + 0.35) = 0.7.

[0164] With this design, whether in the scenario of cardiovascular disease prevention in the elderly or diabetes prevention in obese groups, the server can accurately determine the first knowledge logic density of the first nursing logic derivation knowledge in the first knowledge representation and the first knowledge logic weight in multiple first knowledge representations. These results provide important data support for subsequent more accurate health monitoring and nursing decisions.

[0165] Figure 2 The smart care system 100 shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the smart care system 100 may also include a transceiver 1004, which can be used for data interaction between the server and other servers, such as data transmission and / or data reception. It should be noted that the transceiver 1004 is not limited to one in actual scheduling, and the structure of the smart care system 100 does not constitute a limitation on the embodiments of the present application.

[0166] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0167] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For the convenience of diagram coding features, Figure 2 The feature is encoded using only one thick line, but it does not encode the feature of only one bus or one type of bus.

[0168] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, without limitation herein.

[0169] The memory 1003 is used to store program codes for executing the embodiments of the present application, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the above method embodiments.

[0170] In addition, an embodiment of the present application also provides a readable storage medium, in which computer executable instructions are pre-installed. When the processor executes the computer executable instructions, the page information push method applied to user health management as above is implemented.

[0171] Similarly, it should be noted that in order to simplify the description disclosed in this application, thereby helping to understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof. Similarly, it should be noted that in order to simplify the description disclosed in this application, thereby helping to understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present application, multiple features are sometimes combined into one embodiment, drawings, or descriptions thereof.

Claims

1. A page information push method applied to user health management, characterized in that: The method comprises: Integrate and output the knowledge node data corresponding to the same nursing migration stage in the user nursing health knowledge graph data as first knowledge node path data, wherein the user nursing health knowledge graph data is a data set to be analyzed and personalized nursing suggestions to be pushed; Integrate and output the first knowledge node path data corresponding to the same nursing knowledge topic in the user nursing health knowledge graph data as second knowledge node path data; Loading the second knowledge node path data into a graph neural network model to generate a first graph encoding feature for expressing the second knowledge node path data; Acquire personalized nursing knowledge point data, determine a first feature distance between the personalized nursing knowledge point data and the first graph encoding feature of each of the second knowledge node path data, and determine key knowledge node path data among a plurality of the second knowledge node path data based on the first feature distance; Pushing nursing execution suggestion information matching the personalized nursing knowledge point data for the subsequent nursing stage corresponding to the key knowledge node path data; The step of loading the second knowledge node path data into a graph neural network model and generating a first graph encoding feature for expressing the second knowledge node path data includes: Acquire collaborative annotation information corresponding to the second knowledge node path data from the user care health knowledge graph data; Loading the second knowledge node path data and the collaborative annotation information into the graph neural network model simultaneously, generating the first graph encoding feature for expressing the second knowledge node path data; The acquiring, from the user care health knowledge graph data, collaborative annotation information corresponding to the second knowledge node path data comprises: Extracting nursing operation information from the second knowledge node path data through an entity recognition network and a relationship extraction network; Extracting user nursing profile information corresponding to the nursing operation information from the second knowledge node path data through a user profile analysis network; The nursing operation information and the user nursing portrait information corresponding to the nursing operation information are aggregated to generate the collaborative annotation information.

2. The page information push method for user health management according to claim 1, characterized in that: The step of extracting nursing operation information from the second knowledge node path data through the entity recognition network and the relationship extraction network includes: Identify various structural elements in the second knowledge node path data through an entity recognition network, and after formatting and converting the various structural elements, perform preliminary division of the formatted conversion data into semantic units according to a professional vocabulary library and semantic segmentation rules in the nursing field to generate different semantic units, wherein the semantic units are single words, phrases or sentence structures, and the various structural elements include different levels of labels, indexes and existing metadata information; Pre-construct a corpus, and use the corpus to pre-label each semantic unit in the second knowledge node path data, wherein the corpus contains text materials related to pre-nursing operations, and the pre-labeling process is achieved by calculating the similarity between each semantic unit and different semantic categories in the corpus, wherein for each semantic unit, after converting the semantic unit into a representation form of a feature vector through a word vector model, the similarity between the feature vector of the semantic unit and the different semantic category vectors pre-labeled in the implementation corpus is calculated, and each semantic unit is pre-labeled as a corresponding semantic category according to the similarity; For each semantic unit with a pre-labeled semantic category, the context environment of the semantic unit is analyzed, and the context environment of the semantic unit is perceived through a pre-built context semantic model to calculate the dynamic correlation between the semantic unit and different semantic categories in the current context, and according to the dynamic correlation between the semantic unit and different semantic categories in the current context, the context environment includes the previous and next semantic units of the semantic unit and the semantic categories of the previous and next semantic units, and the context semantic model is implemented based on a neural network or a semantic analysis model based on a graph structure; According to the dynamic correlation between the semantic unit and different semantic categories in the current context, the semantic category pre-annotated for the semantic unit is adjusted, and after the second knowledge node path data of the re-annotated semantic category is generated, the potential entity recognition rule is used to identify the potential entities in the second knowledge node path data of the re-annotated semantic category, wherein the potential entities refer to various objects related to the nursing operation, including the executor of the nursing operation, the object of the nursing operation, the tool used for the nursing operation, and the relevant environmental elements of the nursing operation; After identifying the potential entities, based on the semantic similarity of the potential entities and the relevance of the potential entities in the nursing operation, the potential entities are preliminarily clustered to generate at least one entity cluster; Analyzing the grammatical structure and semantic logical relationship in the second knowledge node path data of the re-annotated semantic category through a relationship extraction network, and identifying the relationship pattern between the entities in the at least one entity cluster according to the grammatical structure and semantic logical relationship, wherein the relationship pattern includes a subject-predicate relationship pattern, a verb-object relationship pattern, or a modifier-modifier relationship pattern; For each of the entity clusters, construct corresponding relationship candidate information based on the identified relationship pattern, wherein the relationship candidate information is an expression describing the relationship between entities; A nursing operation logic knowledge base is established in advance, and the nursing operation logic knowledge base contains nursing operation logic rules in the nursing field. For each combination data of the entity cluster and the relationship candidate information, a verification and adjustment process is performed according to the nursing operation logic knowledge base, and nursing operation information is extracted from the adjusted combination data. The nursing operation information includes the specific name of the nursing operation, the executor of the operation, the object of the operation, the tool used for the operation, and the relevant conditions of the operation.

3. The page information push method for user health management according to claim 1, characterized in that: The step of loading the second knowledge node path data and the collaborative annotation information into the graph neural network model simultaneously to generate the first graph encoding feature for expressing the second knowledge node path data includes: Fusion of the collaborative annotation information into the first guidance information to generate first fused annotation information, wherein the first guidance information includes metadata, configuration information or indicative data related to the processing of the graph neural network model, and is used to enhance the graph neural network model's ability to understand and encode the knowledge node path data; The second knowledge node path data and the first fused annotation information are simultaneously loaded into the graph neural network model to generate the first graph encoding feature for expressing the second knowledge node path data.

4. The page information push method for user health management according to claim 3, characterized in that: The step of loading the second knowledge node path data and the first fusion annotation information into the graph neural network model simultaneously to generate the first graph encoding feature for expressing the second knowledge node path data includes: Extracting a plurality of first focused knowledge node data from the second knowledge node path data, and generating a first focused knowledge node vector according to the first focused knowledge node data; Editing a plurality of the first focused knowledge node vectors into a first input feature representation for loading into the graph neural network model; Generating a second input feature representation for loading into the graph neural network model based on the first fused annotation information; The first input feature representation and the second input feature representation are simultaneously loaded into the graph neural network model to generate the first graph encoding feature for expressing the second knowledge node path data.

5. The page information push method for user health management according to claim 1, characterized in that: The first graph encoding feature is a first knowledge representation of the second knowledge node path data; The determining of the first feature distance between the first graph encoding feature of the personalized nursing knowledge point data and each of the second knowledge node path data comprises: Extracting a plurality of first nursing logic derivation knowledge from the personalized nursing knowledge point data; For each piece of the first nursing logic derived knowledge, determining a first matching degree between the first nursing logic derived knowledge and the first knowledge representation; The first characteristic distance between the personalized nursing knowledge point data and the first knowledge representation is determined according to the first matching degree between each of the first nursing logic-derived knowledge and the first knowledge representation.

6. The page information push method for user health management according to claim 5, characterized in that: The step of extracting a plurality of first nursing logic derivation knowledge from the personalized nursing knowledge point data comprises: Pre-constructing a semantic analysis framework, which is constructed based on the professional knowledge system in the nursing field and includes definitions and descriptions of various concepts, entities, and relationship semantic elements in the nursing field; Dividing the personalized nursing knowledge point data into semantic regions according to the semantic analysis framework to generate data of different semantic regions or semantic relationship parts; Decomposing the data in each semantic region or semantic relationship part into a plurality of semantic units, analyzing the semantic connotation and semantic boundary of each semantic unit, determining the semantic features of the semantic unit in the nursing semantic system, and analyzing the internal relationships between the semantic units based on the semantic features of the semantic units in the nursing semantic system, and establishing a mapping relationship between the semantic units; Based on the mapping relationship between the semantic units, the actual existing logical relationship instances are searched and identified in the personalized nursing knowledge point data, and for each logical relationship instance, the semantic units involved in the logical relationship instance are determined, and the logical relationship types between the semantic units in the logical relationship instance are constructed, wherein in the process of constructing the logical relationship types, the logical relationships across semantic areas are integrated; Pre-constructing a candidate set of logical deduction knowledge patterns, wherein each logical deduction knowledge pattern in the candidate set of logical deduction knowledge patterns includes corresponding logical elements and logical relationship types between the logical elements; Matching and associating the logical relationship type between each of the semantic units in the logical relationship instance with the candidate logical deduction knowledge pattern set, and for each logical relationship instance, searching for a matching target logical deduction knowledge pattern in the candidate logical deduction knowledge pattern set, wherein the matching process is achieved by comparing the logical element type, logical relationship type and global logical structure in the logical relationship instance with the association between the corresponding parts in the logical deduction knowledge pattern; According to the target logic deduction knowledge pattern matched by each logical relationship instance, nursing logic deduction knowledge is extracted from the personalized nursing knowledge point data, wherein, for each logical relationship instance marked with the target logic deduction knowledge pattern, the corresponding nursing logic deduction knowledge is extracted from the logical relationship instance according to the logical element definition in the target logic deduction knowledge pattern.

7. The page information push method for user health management according to claim 5, characterized in that: The determining, for each of the first nursing logic derived knowledge, a first matching degree between the first nursing logic derived knowledge and the first knowledge representation includes: For each of the first nursing logic derived knowledge, determining a first knowledge logic density of the first nursing logic derived knowledge in the first knowledge representation; determining a first knowledge logic weight of the first nursing logic-derived knowledge in a plurality of the first knowledge representations; Determining the first matching degree according to the first knowledge logic density and the first knowledge logic weight; Wherein, the step of determining the first knowledge logic density of the first nursing logic derived knowledge in the first knowledge representation for each of the first nursing logic derived knowledge includes; Decomposing the first knowledge representation into a plurality of semantic units, and marking the semantic type, position information and semantic boundary information of each semantic unit; Decomposing the first nursing logic derivation knowledge into multiple components, where different components are different semantic concepts, logical relationship elements or knowledge fragments; Determine the semantic role of each of the components, the logical connection mode between the components, and the semantic boundaries and semantic connotations between the components, and establish an association relationship between each component and the global semantic structure of the first nursing logic derivation knowledge, wherein the semantic role includes a logical starting point, a logical intermediate link, and a logical result part, and the logical connection mode includes a causal relationship, a progressive relationship, and a parallel relationship; By comparing the similarity of semantic connotations, semantic types and logical roles, a semantic mapping relationship between each component of the first nursing logic derivation knowledge and the semantic unit in the first knowledge representation is constructed, and based on the semantic mapping relationship, a target area that is logically associated with the first nursing logic derivation knowledge is identified in the first knowledge representation, wherein, for each semantic unit in the first knowledge representation in the mapping relationship, a target area is identified with the semantic unit as the center according to the logical connection relationship between the semantic units, and the target area contains other semantic units that are logically related to the semantic unit; Retrieving a pre-built logic density calculation framework, wherein the logic density calculation framework includes evaluation indicators of multiple dimensions, calculation rules and weight allocation rules for each evaluation indicator, and the relationship between each evaluation indicator, wherein the evaluation indicators include the number of semantic units, the richness of the logical relationship between the semantic units, and the semantic matching degree between the logical association area and the first nursing logic derivation knowledge as a whole; According to the logic density calculation framework, the logic density of the first nursing logic derivation knowledge in the first knowledge representation is calculated, wherein the evaluation value of the target area on each evaluation indicator is calculated respectively according to the calculation rules of each evaluation indicator, and the evaluation value of each evaluation indicator is weighted according to the weight distribution principle to obtain a preliminary logic density value, and the preliminary logic density value is adjusted according to the relationship between each evaluation indicator to obtain the final first knowledge logic density; The step of determining the first knowledge logic weight of the first nursing logic derivation knowledge in the plurality of first knowledge representations includes: Performing semantic feature analysis on multiple first knowledge representations to obtain semantic features in each first knowledge representation, and recording distribution parameters of each semantic feature in each first knowledge representation, wherein the semantic features are determined by analyzing key semantic units in the first knowledge representation, logical relationships between the key semantic units, and a global semantic structure, and the distribution parameters include a proportion of each semantic feature in a first knowledge representation, and a type of logical relationship between the semantic feature and other semantic features; Decomposing the semantic content of the first nursing logic derivation knowledge into core semantic elements and logical relationship parts, and determining the semantic connotation, semantic boundary and semantic role of each core semantic element in the nursing logic derivation knowledge, as well as the logical relationship type between each core semantic element; For each first knowledge representation, the semantic association between the first nursing logic derivation knowledge and the first knowledge representation is evaluated, wherein the core semantic elements of the first nursing logic derivation knowledge are specifically compared with the semantic features in the first knowledge representation. For each core semantic element, by comparing the semantic connotation, semantic type and semantic role of the core semantic element, a target semantic feature semantically related to the core semantic element is found in the first knowledge representation, and after determining the semantic association score corresponding to the core semantic element and the corresponding target semantic feature, the semantic association score is weightedly calculated in combination with the matching degree of the logical relationship type between each core semantic element and the logical relationship type between each semantic feature in the first knowledge representation, so as to obtain the semantic association between the first nursing logic derivation knowledge and each first knowledge representation; The obtained semantic association is normalized to obtain the first knowledge logic weight of the first nursing logic derivation knowledge in the first knowledge representation.

8. A smart nursing system, characterized in that: The smart care system includes a processor and a machine-readable storage medium, wherein the machine-readable storage medium stores machine-executable instructions, and the machine-executable instructions are loaded and executed by the processor to implement the page information push method for user health management described in any one of claims 1-7.

Images