Diabetes intervention intelligent recommendation method based on knowledge graph

Through deep learning technology, the problem of semantic embedding and encoding of diabetes response strategies has been solved, the problem of semantic information omission in the existing technology has been improved, the accuracy and comprehensiveness of diabetes intervention plans have been improved, and the construction of personalized intervention plans has been realized.

CN119092036AInactive Publication Date: 2024-12-06ZHEJIANG CHINESE MEDICAL UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411127446.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing intelligent recommendation method for diabetes intervention based on knowledge graph ignores semantic information when coping with strategy query matching, resulting in omissions in the construction of intervention solution library and the inability to achieve personalized and precise intervention.

Method used

Using artificial intelligence technology based on deep learning, the target word and alternative coping strategies are semantic embedding and semantic encoding, and through fine-grained semantic correlation matching, the inclusion of coping strategies is determined, and the word embedding matrix, Bert model, and energy are used. Level-driven semantic correlation radiation enhancement module and adaptive fine-grained feature decoupling query response module to capture semantic features and correlation.

Benefits of technology

It effectively compensates for the problem of semantic information omission in keyword matching technology, improves the accuracy and comprehensiveness of diabetes intervention plans, and ensures the personalization and accuracy of intervention plans.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119092036A_ABST
    Figure CN119092036A_ABST
Patent Text Reader

Abstract

The invention relates to a diabetes intervention intelligent recommendation method based on a knowledge graph. The method comprises the following steps: respectively performing semantic analysis on a target word and an alternative coping strategy by adopting an artificial intelligence technology based on deep learning, respectively capturing semantic feature representations of the target word and the alternative coping strategy, and further performing fine-grained deep semantic association matching on the target word and the alternative coping strategy, and intelligently judging whether the alternative coping strategy is brought into the intervention scheme library or not. Therefore, the problem of inaccurate intervention scheme matching caused by semantic information omission in a keyword matching technology-based diabetes disease coping strategy query method in the prior art can be effectively solved, so that the accuracy and comprehensiveness of diabetes intervention scheme construction are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of diabetes intervention, and specifically, to an intelligent recommendation method for diabetes intervention based on a knowledge graph. Background Art

[0002] In recent years, with the change of lifestyle, the incidence of diabetes has shown an upward trend worldwide. Diabetes is a chronic metabolic disease, the main feature of which is insufficient insulin secretion or decreased ability of the body to use insulin, leading to persistently high blood sugar levels. The long-term presence of diabetes can lead to a variety of complications, such as cardiovascular disease, nephropathy, retinopathy, etc., which seriously affect the patient's quality of life and life expectancy. Therefore, early intervention and effective management of diabetes are particularly important.

[0003] Traditional diabetes management mainly relies on the professional guidance of doctors and the self-management of patients, but due to limited doctor resources and differences in patients' knowledge level and self-management ability, this method is often difficult to achieve individualized and precise intervention. In this regard, in the Chinese invention application with application publication number CN116013511A, a diabetes intervention intelligent recommendation method and system based on knowledge graph is disclosed, which constructs a graph data set by retrieving diabetes-related data, collects patient symptom characteristics through an automatic consultation system, selects conventional coping strategies from the knowledge graph based on the patient's symptom characteristics, and verifies its feasibility through simulation analysis. If the conventional solution is not feasible, a symptom feature library is constructed, and the symptom characteristics in the symptom feature library are used as target words. One or more corresponding coping strategies are queried and matched from the knowledge graph, and an intervention solution library is constructed to achieve intelligent recommendation of diabetes intervention solutions.

[0004] However, when the above scheme matches the query of coping strategies based on the target word, it ignores the contextual semantic information in the coping strategy and the semantic association between the target word and the coping strategy because it adopts the query method of keyword matching. It usually misses the relevant semantic information and leads to omissions in the construction of the intervention plan library.

[0005] Therefore, an optimized knowledge graph-based intelligent recommendation method for diabetes intervention is expected. Summary of the invention

[0006] This summary is provided to introduce concepts in a brief form that will be described in detail in the detailed description below. This summary is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0007] In a first aspect, the present application provides a diabetes intervention intelligent recommendation method based on a knowledge graph, the method comprising:

[0008] Get the target word;

[0009] Processing the target word using a word embedding matrix to obtain a semantic embedding vector of the target word;

[0010] Extract the first alternative response strategy from the knowledge graph;

[0011] Performing word-granularity-based semantic coding and semantic association radiation enhancement on the first alternative response strategy to obtain a sequence of word-granularity-enhanced semantic coding feature vectors of the first alternative response strategy;

[0012] Based on the fine-grained feature matching between the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity-enhanced semantic encoding feature vector, determine whether to include the first alternative coping strategy into the intervention plan library.

[0013] Optionally, the first alternative response strategy is subjected to word-granularity-based semantic coding and semantic association radiation enhancement to obtain a sequence of first alternative response strategy word-granularity enhanced semantic coding feature vectors, including: performing word segmentation processing on the first alternative response strategy and then passing it through a semantic encoder including a Bert model to obtain a sequence of first alternative response strategy word-granularity semantic coding feature vectors; performing semantic association radiation enhancement on the sequence of word-granularity semantic coding feature vectors of the first alternative response strategy to obtain a sequence of word-granularity enhanced semantic coding feature vectors of the first alternative response strategy.

[0014] Optionally, the sequence of the first alternative response strategy word granularity semantic coding feature vectors is subjected to semantic association radiation enhancement to obtain the sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors, including: inputting the sequence of the first alternative response strategy word granularity semantic coding feature vectors into a semantic association radiation enhancement module based on energy level drive to obtain the sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors.

[0015] Optionally, the sequence of the first alternative response strategy word granularity semantic coding feature vectors is input into a semantic association radiation enhancement module based on energy level drive to obtain a sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors, including: calculating the characteristic energy level coefficients of each first alternative response strategy word granularity semantic coding feature vector in the sequence of the first alternative response strategy word granularity semantic coding feature vectors to obtain a sequence of characteristic energy level coefficients; arranging the sequence of characteristic energy level coefficients into a characteristic energy level coefficient input vector; performing one-dimensional convolution encoding on the characteristic energy level coefficient input vector to obtain a characteristic energy level radiation association vector; performing point convolution encoding and probabilistic mapping based on Sigmoid function on the characteristic energy level radiation association vector to obtain a characteristic energy level radiation association weight vector; using the characteristic energy level radiation association weight vector as a weight vector, weighting processing and residual processing are performed on the sequence of the first alternative response strategy word granularity semantic coding feature vectors to obtain a sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors.

[0016] Optionally, the characteristic energy level coefficients of each first alternative response strategy word granularity semantic coding feature vector in the sequence of the first alternative response strategy word granularity semantic coding feature vectors are calculated to obtain a sequence of characteristic energy level coefficients, including: calculating the weighted sum between the variance of the first alternative response strategy word granularity semantic coding feature vector and the bias term to obtain a first intermediate term, and calculating the square of the difference between the maximum value and the mean of the first alternative response strategy word granularity semantic coding feature vector, the variance of the first alternative response strategy word granularity semantic coding feature vector and the bias term to obtain a second intermediate term; calculating the first intermediate term divided by the second intermediate term to obtain the characteristic energy level coefficient.

[0017] Optionally, based on the fine-grained feature matching between the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity-enhanced semantic encoding feature vector, determining whether to include the first alternative coping strategy in the intervention plan library includes: performing fine-grained query matching analysis on the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity-enhanced semantic encoding feature vector to obtain a target word-first alternative coping strategy query response adaptive matching representation matrix; inputting the target word-first alternative coping strategy query response adaptive matching representation matrix into a classifier-based pairing result generator to obtain a pairing result, and the pairing result is used to indicate whether to include the first alternative coping strategy.

[0018] Optionally, a fine-grained query matching analysis is performed on the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity-enhanced semantic encoding feature vector to obtain a target word-first alternative response strategy query response adaptive matching representation matrix, including: using the target word semantic embedding vector as a query feature vector, and inputting the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity-enhanced semantic encoding feature vector into an adaptive fine-grained feature decoupling query response module to obtain the target word-first alternative response strategy query response adaptive matching representation matrix.

[0019] Optionally, the target word semantic embedding vector is used as the query feature vector, and the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity enhanced semantic coding feature vector is input into the adaptive fine-grained feature decoupling query response module to obtain the target word-first alternative response strategy query response adaptive matching representation matrix, including: respectively performing local semantic association enhancement on each first alternative response strategy word granularity enhanced semantic coding feature vector in the sequence of the first alternative response strategy word granularity enhanced semantic coding feature vector to obtain a sequence of first alternative response strategy word granularity local semantic association feature vectors; calculating the local semantic association relationship between the query feature vector and the first alternative response strategy word granularity The method comprises the following steps: calculating the covariance matrix between each first alternative response strategy word granularity local semantic association feature vector in the sequence of linked feature vectors to obtain a sequence of query response covariance matrices; calculating the matrix traces of each query response covariance matrix in the sequence of query response covariance matrices to obtain a sequence of query response trace measurement factors; inputting the sequence of query response trace measurement factors into a Sigmoid activation function to obtain a sequence of query response trace attention weights; using each query response trace attention weight in the sequence of query response trace attention weights as a weight, calculating the weighted sum of the sequence of query response covariance matrices to obtain the target word-first alternative response strategy query response adaptive matching representation matrix.

[0020] Optionally, each first alternative response strategy word granularity enhanced semantic coding feature vector in the sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors is respectively subjected to local semantic association enhancement to obtain a sequence of first alternative response strategy word granularity local semantic association feature vectors, including: fine-grained decoupling of the first alternative response strategy word granularity enhanced semantic coding feature vectors to obtain a sequence of first alternative response strategy word granularity local semantic coding feature vectors; inputting the sequence of the first alternative response strategy word granularity local semantic coding feature vectors into a sequence semantic enhancement encoder based on a Bi-LSTM model to obtain the first alternative response strategy word granularity local semantic association feature vector.

[0021] By adopting the above technical solution, the target word and the alternative coping strategy are semantically analyzed by using artificial intelligence technology based on deep learning, and the semantic feature representation of the target word and the alternative coping strategy is captured respectively, and then the target word and the alternative coping strategy are matched with each other in a fine-grained deep semantic association to intelligently determine whether to include the alternative coping strategy in the intervention plan library. In this way, the problem of inaccurate intervention plan matching caused by the omission of semantic information in the diabetes disease coping strategy query method based on keyword matching technology in the prior art can be effectively compensated, thereby improving the accuracy and comprehensiveness of the construction of diabetes intervention plans.

[0022] Other features and advantages of the present application will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The above and other features, advantages and aspects of the embodiments of the present application will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale. In the drawings:

[0024] Figure 1 The present invention is a flowchart of a method for intelligently recommending diabetes intervention based on a knowledge graph according to an exemplary embodiment.

[0025] Figure 2 It is a block diagram of a diabetes intervention intelligent recommendation system based on knowledge graph according to an exemplary embodiment.

[0026] Figure 3 It is a block diagram of an electronic device according to an exemplary embodiment.

[0027] Figure 4 This is an application scenario diagram of a knowledge graph-based intelligent recommendation method for diabetes intervention according to an exemplary embodiment. DETAILED DESCRIPTION

[0028] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present application. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes and are not intended to limit the scope of protection of the present application.

[0029] It should be understood that the various steps described in the method implementation of the present application can be performed in different orders and / or performed in parallel. In addition, the method implementation may include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.

[0030] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.

[0031] It should be noted that the concepts such as "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0032] It should be noted that the modifications of "one" and "plurality" mentioned in the present application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0033] The names of the messages or information exchanged between multiple devices in the embodiments of the present application are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0034] The specific implementation of the present application is described in detail below with reference to the accompanying drawings.

[0035] Figure 1 is a flowchart of a diabetes intervention intelligent recommendation method based on a knowledge graph according to an exemplary embodiment. Figure 1 As shown, the method includes:

[0036] Step S101, obtaining a target word;

[0037] Step S102: Process the target word using a word embedding matrix to obtain a semantic embedding vector of the target word;

[0038] Step S103: extracting a first alternative response strategy from the knowledge graph;

[0039] Step S104, performing word-granularity-based semantic coding and semantic association radiation enhancement on the first candidate response strategy to obtain a sequence of word-granularity-enhanced semantic coding feature vectors of the first candidate response strategy;

[0040] Step S105: Determine whether to include the first alternative coping strategy into the intervention plan library based on the fine-grained feature matching between the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity enhanced semantic encoding feature vector.

[0041] In response to the above technical problems, the technical concept of this application is to use artificial intelligence technology based on deep learning to perform semantic analysis on the target word and the alternative coping strategy, respectively, to capture the semantic feature representation of the target word and the alternative coping strategy, and then to intelligently determine whether to include the alternative coping strategy in the intervention plan library by performing fine-grained deep semantic association matching between the target word and the alternative coping strategy. In this way, it can effectively make up for the problem of inaccurate intervention plan matching caused by the omission of semantic information in the diabetes disease coping strategy query method based on keyword matching technology in the prior art, thereby improving the accuracy and comprehensiveness of the construction of diabetes intervention plans.

[0042] Based on this, in the technical solution of the present application, first, the target words are obtained. It should be understood that the target words are words related to diabetes management, such as specific symptoms, complications, etc. of diabetic patients, such as "high blood sugar", "drinking and eating more", "blurred vision", etc. In the technical solution of the present application, the target words are obtained to clarify the intervention needs, providing accurate positioning for the subsequent matching of response strategies.

[0043] Next, considering that directly using the original form of the target word (such as a string) for matching can often only capture surface similarities and ignore deep semantic associations, in order to obtain the semantic information of the target word and understand the intervention needs, the present application further embeds the target word through a word embedding matrix to capture the potential semantic connotation of the vocabulary, and converts it into a vector form to generate a target word semantic embedding vector for subsequent semantic level analysis and matching.

[0044] At the same time, the first alternative coping strategy is extracted from the knowledge graph. The first alternative coping strategy is a series of coping measures or suggestions for possible symptoms of diabetes in the knowledge graph, which may include various types such as adjusting diet, increasing exercise, and drug treatment. In order to accurately match the coping strategy suitable for the patient, this application will use a more refined analysis method to perform semantic analysis on the first alternative coping strategy. First, the first alternative coping strategy is segmented and decomposed into multiple vocabulary units to facilitate subsequent fine-grained semantic analysis of each vocabulary unit. Then, each vocabulary unit is embedded in coding and contextual semantic association analysis is performed through a semantic encoder including a Bert model. The powerful semantic understanding ability of the Bert model is used to capture the semantic information of each vocabulary unit in a specific context, and generate a sequence of word-granular semantic coding feature vectors of the first alternative coping strategy.

[0045] In one embodiment of the present application, the first alternative response strategy is subjected to word-granularity-based semantic coding and semantic association radiation enhancement to obtain a sequence of first alternative response strategy word-granularity enhanced semantic coding feature vectors, including: performing word segmentation processing on the first alternative response strategy and then passing it through a semantic encoder including a Bert model to obtain a sequence of first alternative response strategy word-granularity semantic coding feature vectors; performing semantic association radiation enhancement on the sequence of first alternative response strategy word-granularity semantic coding feature vectors to obtain a sequence of first alternative response strategy word-granularity enhanced semantic coding feature vectors.

[0046] Furthermore, considering that in the first alternative response strategy, different vocabulary units may have different semantic contributions to the response strategy, some vocabulary units may be more core or critical, while others may be relatively minor. In order to further improve the feature expression ability of the sequence of the word-granular semantic coding feature vectors of the first alternative response strategy, the present application further introduces a semantic association radiation enhancement module based on energy level drive to process the sequence of the word-granular semantic coding feature vectors of the first alternative response strategy. Specifically, the semantic association radiation enhancement module quantifies the semantic feature information density of each vocabulary unit and its semantic influence on other vocabulary units in the surrounding neighborhood by calculating the semantic feature energy level coefficient of each vocabulary unit, and captures the local semantic dependency between vocabulary by performing a one-dimensional convolution operation on each feature energy level coefficient, so as to generate attention weights, and perform weighted processing and residual processing on the sequence of the word-granular semantic coding feature vectors of the first alternative response strategy, thereby enhancing the semantic influence of the core vocabulary by evaluating the differences in semantic contributions between vocabulary units, and generating a sequence of enhanced semantic coding feature vectors of the word-granularity of the first alternative response strategy.

[0047] In one embodiment of the present application, a sequence of the first alternative response strategy word granularity semantic coding feature vectors is subjected to semantic association radiation enhancement to obtain a sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors, including: inputting the sequence of the first alternative response strategy word granularity semantic coding feature vectors into a semantic association radiation enhancement module based on energy level drive to obtain a sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors.

[0048] Furthermore, in one embodiment of the present application, the sequence of the first alternative response strategy word granularity semantic coding feature vectors is input into a semantic association radiation enhancement module based on energy level drive to obtain a sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors, including: calculating the characteristic energy level coefficients of each first alternative response strategy word granularity semantic coding feature vector in the sequence of the first alternative response strategy word granularity semantic coding feature vectors to obtain a sequence of characteristic energy level coefficients; arranging the sequence of characteristic energy level coefficients into a characteristic energy level coefficient input vector; performing one-dimensional convolution encoding on the characteristic energy level coefficient input vector to obtain a characteristic energy level radiation association vector; performing point convolution encoding and probabilistic mapping based on the Sigmoid function on the characteristic energy level radiation association vector to obtain a characteristic energy level radiation association weight vector; using the characteristic energy level radiation association weight vector as a weight vector, weighting processing and residual processing are performed on the sequence of the first alternative response strategy word granularity semantic coding feature vectors to obtain a sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors.

[0049] Furthermore, in one embodiment of the present application, the characteristic energy level coefficients of each first alternative response strategy word granularity semantic coding feature vector in the sequence of the first alternative response strategy word granularity semantic coding feature vectors are calculated to obtain a sequence of characteristic energy level coefficients, including: calculating the weighted sum between the variance of the first alternative response strategy word granularity semantic coding feature vector and the bias term to obtain a first intermediate term, and calculating the square of the difference between the maximum value and the mean of the first alternative response strategy word granularity semantic coding feature vector, the variance of the first alternative response strategy word granularity semantic coding feature vector and the bias term to obtain a second intermediate term; calculating the first intermediate term divided by the second intermediate term to obtain the characteristic energy level coefficient.

[0050] Specifically, the sequence of the first candidate response strategy word granularity semantic coding feature vectors is processed by the following semantic association enhancement formula to obtain the sequence of the first candidate response strategy word granularity enhanced semantic coding feature vectors, wherein the semantic association enhancement formula is:

[0051] X={x1,x2,...,x i ,...,x n}

[0052]

[0053] V e =(e1; e2; ...; e i ;...;e n )

[0054] V f =Conv1×l (V e )

[0055] V c =Sigmoid(Conv 1×1 (V f ))

[0056] X′=X⊙V c +X

[0057] Wherein, X represents the sequence of the first candidate response strategy word granular semantic coding feature vectors, the value of n is the number of feature vectors in the sequence of the first candidate response strategy word granular semantic coding feature vectors, x1, x2, x i and x n Respectively represent the first, second, i-th and n-th first candidate response strategy word granularity semantic encoding feature vectors in the sequence of the first candidate response strategy word granularity semantic encoding feature vectors, μ i is the feature mean of the granular semantic encoding feature vector of the i-th first alternative response strategy word, σ 2 i is the feature variance of the semantic encoding feature vector of the i-th first candidate response strategy word, max(x i ) represents the maximum value of each feature value in the granular semantic encoding feature vector of the i-th first candidate response strategy word, ε is the bias term, e1, e2, e i and e n Respectively represent the characteristic energy level coefficients of the first, second, i-th and n-th first candidate response strategy word granular semantic coding feature vectors in the sequence of the first candidate response strategy word granular semantic coding feature vectors, V e Represents the characteristic energy level coefficient input vector, Conv 1×l (·) represents a one-dimensional convolution operation, l is the scale of the one-dimensional convolution kernel, V f Represents the characteristic energy level radiation correlation vector, Conv 1×1 represents the point convolution operation, Sigmoid represents the Sigmoid function, V c represents the characteristic energy level radiation association weight vector, ⊙ represents the dot product, and X′ represents the sequence of the first alternative response strategy word granularity enhanced semantic encoding feature vectors.

[0058] Next, a matching analysis is further performed on the target word semantic embedding vector and the sequence of the first alternative response strategy word granularity-enhanced semantic coding feature vector to determine the degree of adaptation between the target word and the first alternative response strategy. In the technical solution of the present application, an adaptive fine-grained feature decoupling query response module is introduced to perform fine-grained matching analysis and processing on the two. Specifically, the adaptive fine-grained feature decoupling query response module first performs fine-grained decoupling on each first alternative response strategy word granularity-enhanced semantic coding feature vector, splits it into a sequence of local feature vectors, and uses a Bi-LSTM model to perform local information semantic association encoding on it to enhance its local semantic information expression capability, so that each first alternative response strategy word granularity-enhanced semantic coding feature vector has stronger expression capability and association at the local level. Next, the covariance matrix between the target word semantic embedding vector and the first alternative coping strategy word granular enhanced semantic encoding feature vector after each local semantic association encoding and the trace of the covariance matrix are calculated to quantify the correlation between the two, and the attention weight is generated. By weighting and calculating each covariance matrix, the target word-first alternative coping strategy query response adaptive matching representation matrix is ​​obtained, which comprehensively represents the fine-grained matching degree and semantic association strength between the target word and the first alternative coping strategy, providing more accurate data support for the subsequent construction of intervention plans.

[0059] In one embodiment of the present application, based on the fine-grained feature matching between the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity-enhanced semantic encoding feature vector, determining whether to include the first alternative coping strategy in the intervention plan library includes: performing fine-grained query matching analysis on the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity-enhanced semantic encoding feature vector to obtain a target word-first alternative coping strategy query response adaptive matching representation matrix; inputting the target word-first alternative coping strategy query response adaptive matching representation matrix into a classifier-based pairing result generator to obtain a pairing result, and the pairing result is used to indicate whether to include the first alternative coping strategy.

[0060] Furthermore, in one embodiment of the present application, a fine-grained query matching analysis is performed on the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity-enhanced semantic encoding feature vector to obtain a target word-first alternative response strategy query response adaptive matching representation matrix, including: using the target word semantic embedding vector as a query feature vector, and inputting the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity-enhanced semantic encoding feature vector into an adaptive fine-grained feature decoupling query response module to obtain the target word-first alternative response strategy query response adaptive matching representation matrix.

[0061] Furthermore, in one embodiment of the present application, the target word semantic embedding vector is used as a query feature vector, and the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity enhanced semantic coding feature vector is input into an adaptive fine-grained feature decoupling query response module to obtain the target word-first alternative response strategy query response adaptive matching representation matrix, including: performing local semantic association enhancement on each first alternative response strategy word granularity enhanced semantic coding feature vector in the sequence of the first alternative response strategy word granularity enhanced semantic coding feature vector to obtain a sequence of first alternative response strategy word granularity local semantic association feature vectors; calculating the query feature vector and the first alternative response strategy word granularity enhanced semantic coding feature vector. The covariance matrix between each first alternative response strategy word granular local semantic association feature vector in the sequence of granular local semantic association feature vectors is used to obtain a sequence of query response covariance matrices; the matrix traces of each query response covariance matrix in the sequence of query response covariance matrices are calculated to obtain a sequence of query response trace measurement factors; the sequence of query response trace measurement factors is input into a Sigmoid activation function to obtain a sequence of query response trace attention weights; and the weighted sum of the sequence of query response covariance matrices is calculated using each query response trace attention weight in the sequence of query response trace attention weights as a weight to obtain the target word-first alternative response strategy query response adaptive matching representation matrix.

[0062] Furthermore, in one embodiment of the present application, each first alternative response strategy word granularity enhanced semantic coding feature vector in the sequence of the first alternative response strategy word granularity enhanced semantic coding feature vectors is respectively subjected to local semantic association enhancement to obtain a sequence of first alternative response strategy word granularity local semantic association feature vectors, including: fine-grained decoupling of the first alternative response strategy word granularity enhanced semantic coding feature vectors to obtain a sequence of first alternative response strategy word granularity local semantic coding feature vectors; inputting the sequence of the first alternative response strategy word granularity local semantic coding feature vectors into a sequence semantic enhancement encoder based on a Bi-LSTM model to obtain the first alternative response strategy word granularity local semantic association feature vector.

[0063] Specifically, the query feature vector and the sequence of the first candidate response strategy word granularity enhanced semantic encoding feature vector are processed by the following query response encoding formula to obtain the target word-first candidate response strategy query response adaptive matching representation matrix:

[0064] S={S1,S2,...,S n}

[0065] S′ i =Decompose(Si )

[0066] {E i}=Bi_LSTM({S′ i})

[0067] {C i}={Cov(Q,E i )}

[0068] {τ i}={Tr(C i )}

[0069] {α i}={σ(τ i )}

[0070]

[0071] Among them, S represents the sequence of the first candidate response strategy word granularity enhanced semantic encoding feature vectors, S1, S2, S i and S n represents the first, second, i-th and n-th first alternative response strategy word granularity reinforcement semantic coding feature vectors in the sequence of the first alternative response strategy word granularity reinforcement semantic coding feature vectors respectively, the value of n is the number of first alternative response strategy word granularity reinforcement semantic coding feature vectors in the sequence of the first alternative response strategy word granularity reinforcement semantic coding feature vectors, Decompose(·) represents the decoupling process, S′ i represents the sequence of local semantic encoding feature vectors of the i-th first alternative response strategy word granularity, Bi_LSTM is a sequence semantic reinforcement encoder based on the Bi-LSTM model, and E i represents the local semantic association feature vector of the i-th first candidate response strategy word granularity, Q represents the query feature vector, Cov(·) represents the covariance calculation function, C i represents the i-th query response covariance matrix, Tr(·) represents the matrix trace calculation function, τ i represents the metric factor of the i-th query response trace, σ represents the Sigmoid activation function, α i represents the attention weight of the i-th query response trace, ⊙ represents the dot product, and m represents the target word-first alternative response strategy query response adaptive matching representation matrix.

[0072] The target word-first alternative coping strategy query response adaptive matching representation matrix is ​​input into the classifier-based pairing result generator to obtain a pairing result, and the pairing result is used to indicate whether to include the first alternative coping strategy. In the technical solution of the present application, the classifier-based pairing result generator has learned the matching of various diabetes-related symptoms (target words) and corresponding coping strategies through repeated training and iterative learning, and can accurately distinguish the adaptability between coping strategies and diabetes symptoms. Furthermore, when the classifier-based pairing result generator receives the target word-first alternative coping strategy query response adaptive matching representation matrix, by using its internally trained weights and parameters, it performs multi-layer nonlinear transformation and feature extraction on it, and can accurately generate a pairing result between the target word and the first alternative coping strategy, and clearly indicate whether the first alternative coping strategy should be included in the personalized intervention program library for diabetic patients.

[0073] In a preferred example, considering that the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity enhanced semantic encoding feature vector respectively represent the text embedding semantic features of the target word and the word granularity context semantic association features of the first alternative response strategy determined based on word segmentation, in this way, when the target word semantic embedding vector is used as the query feature vector and the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity enhanced semantic encoding feature vector are input into the adaptive fine-grained feature decoupling query response module, the target word-first alternative response strategy query response adaptive matching representation matrix obtained will also cause query response matching inconsistency due to the difference in adaptive structural alignment of text feature distribution at text fine granularity, thereby causing distribution regression mode shift, resulting in overflow of the pairing results obtained by the classifier-based pairing result generator of the target word-first alternative response strategy query response adaptive matching representation matrix, affecting its accuracy.

[0074] Based on this, the target word-first candidate response strategy query response adaptive matching representation matrix is ​​passed through a classifier-based pairing result generator to obtain a pairing result, which specifically includes:

[0075] The target word-first candidate response strategy query response adaptive matching representation matrix is ​​passed through a probabilistic activation function to obtain a probabilistic target word-first candidate response strategy query response adaptive matching representation matrix;

[0076] determining an inclusion probability value representing inclusion of the first alternative response strategy;

[0077] After performing point subtraction on the unit feature matrix and the probabilistic target word-first alternative response strategy query response adaptive matching representation matrix, the power function with the inclusion probability value as the base is calculated, and further multiplied with the probabilistic target word-first alternative response strategy query response adaptive matching representation matrix to obtain the probabilistic target word-first alternative response strategy query response adaptive matching representation convergence limiting feature matrix;

[0078] Taking the absolute value of each eigenvalue of the probabilistic target word-first alternative response strategy query response adaptive matching representation matrix as the base, calculating a power function with the inverse of the inclusion probability value as the exponent, and summing all eigenvalues ​​of the probabilistic target word-first alternative response strategy query response adaptive matching representation matrix to obtain a target word-first alternative response strategy query response adaptive matching representation convergence bias eigenvalue;

[0079] The optimized target word-first candidate response strategy query response adaptive matching representation convergence limiting feature matrix and the product of the target word-first candidate response strategy query response adaptive matching representation convergence bias feature value and the weight as a hyperparameter are added to obtain the target word-first candidate response strategy query response adaptive matching representation matrix after optimization; and

[0080] The optimized target word-first candidate coping strategy query response adaptive matching representation matrix is ​​passed through the classifier-based pairing result generator to obtain the pairing result.

[0081] Wherein, let the probabilistic target word-first candidate response strategy query response adaptive matching representation matrix be denoted as M, and the inclusion probability value be p, then the optimized target word-first candidate response strategy query response adaptive matching representation matrix M' is expressed as:

[0082]

[0083] Wherein, M represents the probabilistic target word-first candidate response strategy query response adaptive matching representation matrix obtained after the target word-first candidate response strategy query response adaptive matching representation matrix is ​​activated by the probabilistic function, p represents the inclusion probability value, M I represents the unit feature matrix, m i represents the i-th eigenvalue of the probabilistic target word-first alternative response strategy query response adaptive matching representation matrix, Indicates point addition, represents dot subtraction, ⊙ represents dot multiplication, α represents the weight as a hyperparameter, and M' represents the optimized target word-first alternative response strategy query response adaptive matching representation matrix.

[0084] That is, for the target word-first alternative response strategy query response adaptive matching representation matrix, based on the convergence limit of the Cauchy-Hadamard form power series probability distribution, the feature set of the target word-first alternative response strategy query response adaptive matching representation matrix is ​​represented by a formal power series based on the convergence probability, and a set convergence bounded bias represented by the convergence radius based on each eigenvalue of the target word-first alternative response strategy query response adaptive matching representation matrix is ​​added, so as to avoid the invalid overflow of the classification result caused by the deviation of the distribution pattern of the target word-first alternative response strategy query response adaptive matching representation matrix due to the gradient propagation of the model parameters based on the convergence probability during the regression process of the pairing result generator based on the classifier under a unified probability convergence limit standard, thereby improving the accuracy of the pairing result obtained by the pairing result generator based on the classifier.

[0085] In summary, the above scheme is adopted to perform semantic analysis on the target word and the alternative coping strategy respectively by using artificial intelligence technology based on deep learning, and the semantic feature representation of the target word and the alternative coping strategy is captured respectively, and then the target word and the alternative coping strategy are matched with fine-grained deep semantic associations to intelligently determine whether to include the alternative coping strategy in the intervention plan library. In this way, the problem of inaccurate intervention plan matching caused by the omission of semantic information in the diabetes disease coping strategy query method based on keyword matching technology in the prior art can be effectively compensated, thereby improving the accuracy and comprehensiveness of the construction of diabetes intervention plans.

[0086] Figure 2 FIG. 1 is a block diagram of a diabetes intervention intelligent recommendation system based on a knowledge graph according to an exemplary embodiment. Figure 2 As shown, the system 200 includes:

[0087] A target word acquisition module 201 is used to acquire a target word;

[0088] A target word processing module 202 is used to process the target word using a word embedding matrix to obtain a target word semantic embedding vector;

[0089] A first candidate response strategy extraction module 203, used to extract a first candidate response strategy from the knowledge graph;

[0090] A semantic coding and semantic association radiation enhancement module 204 is used to perform word granularity-based semantic coding and semantic association radiation enhancement on the first candidate response strategy to obtain a sequence of word granularity-enhanced semantic coding feature vectors of the first candidate response strategy;

[0091] The first alternative coping strategy inclusion in the intervention solution library judgment module 205 is used to determine whether to include the first alternative coping strategy in the intervention solution library based on the fine-grained feature matching between the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity enhanced semantic encoding feature vector.

[0092] Reference below Figure 3 , which shows a schematic diagram of the structure of an electronic device 600 suitable for implementing the embodiment of the present application. The terminal device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0093] like Figure 3 As shown, the electronic device 600 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0094] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 3 The electronic device 600 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0095] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present application are executed.

[0096] It should be noted that the computer-readable medium mentioned above in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0097] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0098] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0099] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0100] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0101] The modules involved in the embodiments described in this application can be implemented by software or hardware. The name of the module does not limit the module itself in some cases. For example, the test parameter acquisition module can also be described as a "module for acquiring device test parameters corresponding to the target device".

[0102] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0103] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0104] Figure 4 FIG. 1 is an application scenario diagram of an intelligent recommendation method for diabetes intervention based on a knowledge graph according to an exemplary embodiment. Figure 4 As shown, in this application scenario, first, obtain the target word (for example, Figure 4 C1 as shown in ); and extracting the first alternative response strategy from the knowledge graph (e.g., Figure 4 Then, the obtained target word and the first alternative coping strategy are input into a server (for example, Figure 4 In S) shown in , the server is capable of processing the target word and the first alternative coping strategy based on the diabetes intervention intelligent recommendation algorithm of the knowledge graph to determine whether to include the first alternative coping strategy into the intervention plan library.

[0105] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

[0106] In addition, although each operation is described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the application. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0107] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely exemplary forms of implementation. With respect to the apparatus in the above-mentioned embodiments, the specific manner in which each module performs the operation has been described in detail in the embodiments related to the method, and will not be described in detail here.

Claims

1. A diabetes intervention intelligent recommendation method based on knowledge graph, characterized in that: include: Get the target word; Processing the target word using a word embedding matrix to obtain a semantic embedding vector of the target word; Extract the first alternative response strategy from the knowledge graph; Performing word-granularity-based semantic coding and semantic association radiation enhancement on the first alternative response strategy to obtain a sequence of word-granularity-enhanced semantic coding feature vectors of the first alternative response strategy; Based on the fine-grained feature matching between the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity-enhanced semantic encoding feature vector, determine whether to include the first alternative coping strategy into the intervention plan library.

2. The diabetes intervention intelligent recommendation method based on knowledge graph according to claim 1, characterized in that: The first candidate response strategy is subjected to word granularity-based semantic coding and semantic association radiation enhancement to obtain a sequence of word granularity-enhanced semantic coding feature vectors of the first candidate response strategy, including: After word segmentation processing is performed on the first candidate response strategy, a semantic encoder including a Bert model is used to obtain a sequence of word-granular semantic encoding feature vectors of the first candidate response strategy; The sequence of the first candidate response strategy word granularity semantic coding feature vectors is subjected to semantic association radiation enhancement to obtain a sequence of the first candidate response strategy word granularity enhanced semantic coding feature vectors.

3. The diabetes intervention intelligent recommendation method based on knowledge graph according to claim 2, characterized in that: Performing semantic association radiation enhancement on the sequence of the first candidate response strategy word granularity semantic coding feature vectors to obtain a sequence of the first candidate response strategy word granularity enhanced semantic coding feature vectors, including: The sequence of the first candidate response strategy word granularity semantic coding feature vectors is input into the energy level driven semantic association radiation enhancement module to obtain the sequence of the first candidate response strategy word granularity enhanced semantic coding feature vectors.

4. The diabetes intervention intelligent recommendation method based on knowledge graph according to claim 3, characterized in that: Inputting the sequence of the first candidate coping strategy word granularity semantic encoding feature vectors into the semantic association radiation enhancement module based on energy level drive to obtain the sequence of the first candidate coping strategy word granularity enhanced semantic encoding feature vectors, including: Calculating the characteristic energy level coefficients of each first candidate coping strategy word granular semantic coding feature vector in the sequence of the first candidate coping strategy word granular semantic coding feature vectors to obtain a sequence of characteristic energy level coefficients; Arranging the sequence of characteristic energy level coefficients into a characteristic energy level coefficient input vector; Performing one-dimensional convolution coding on the characteristic energy level coefficient input vector to obtain a characteristic energy level radiation correlation vector; Performing point convolution coding and probabilistic mapping based on a Sigmoid function on the characteristic energy level radiation association vector to obtain a characteristic energy level radiation association weight vector; Using the characteristic energy level radiation associated weight vector as the weight vector, weighted processing and residual processing are performed on the sequence of the first candidate response strategy word granularity semantic coding feature vectors to obtain the sequence of the first candidate response strategy word granularity enhanced semantic coding feature vectors.

5. The diabetes intervention intelligent recommendation method based on knowledge graph according to claim 4, characterized in that: Calculating the characteristic energy level coefficients of each first candidate coping strategy word granular semantic coding feature vector in the sequence of the first candidate coping strategy word granular semantic coding feature vectors to obtain a sequence of characteristic energy level coefficients includes: Calculate the weighted sum of the variance of the first candidate response strategy word granular semantic coding feature vector and the bias term to obtain a first intermediate term, and calculate the square of the difference between the maximum value and the mean value of the first candidate response strategy word granular semantic coding feature vector, the variance of the first candidate response strategy word granular semantic coding feature vector and the bias term to obtain a second intermediate term; The first intermediate term divided by the second intermediate term is calculated to obtain the characteristic energy level coefficient.

6. The diabetes intervention intelligent recommendation method based on knowledge graph according to claim 5, characterized in that: Determining whether to include the first alternative coping strategy into the intervention solution library based on fine-grained feature matching between the target word semantic embedding vector and the sequence of the first alternative coping strategy word granularity enhanced semantic encoding feature vector comprises: Performing fine-grained query matching analysis on the target word semantic embedding vector and the sequence of the first candidate response strategy word granular enhanced semantic encoding feature vector to obtain a target word-first candidate response strategy query response adaptive matching representation matrix; The target word-first candidate response strategy query response adaptive matching representation matrix is ​​input into a classifier-based pairing result generator to obtain a pairing result, and the pairing result is used to indicate whether to include the first candidate response strategy.

7. The diabetes intervention intelligent recommendation method based on knowledge graph according to claim 6, characterized in that: Performing fine-grained query matching analysis on the sequence of the target word semantic embedding vector and the first candidate response strategy word granularity enhanced semantic encoding feature vector to obtain a target word-first candidate response strategy query response adaptive matching representation matrix, including: Taking the target word semantic embedding vector as the query feature vector, the sequence of the target word semantic embedding vector and the first alternative response strategy word granularity enhanced semantic encoding feature vector is input into the adaptive fine-grained feature decoupling query response module to obtain the target word-first alternative response strategy query response adaptive matching representation matrix.

8. The diabetes intervention intelligent recommendation method based on knowledge graph according to claim 7, characterized in that: The target word semantic embedding vector is used as a query feature vector, and a sequence of the target word semantic embedding vector and the first candidate response strategy word granularity enhanced semantic encoding feature vector is input into an adaptive fine-grained feature decoupling query response module to obtain the target word-first candidate response strategy query response adaptive matching representation matrix, including: Performing local semantic association enhancement on each first candidate coping strategy word granularity enhanced semantic coding feature vector in the sequence of the first candidate coping strategy word granularity enhanced semantic coding feature vectors to obtain a sequence of first candidate coping strategy word granularity local semantic association feature vectors; Calculating the covariance matrix between the query feature vector and each first candidate response strategy word granularity local semantic association feature vector in the sequence of the first candidate response strategy word granularity local semantic association feature vectors to obtain a sequence of query response covariance matrices; Calculating the matrix trace of each query response covariance matrix in the sequence of query response covariance matrices to obtain a sequence of query response trace measurement factors; Inputting the sequence of query response trace measurement factors into a sigmoid activation function to obtain a sequence of query response trace attention weights; Taking each query response trace attention weight in the sequence of query response trace attention weights as a weight, a weighted sum of the sequence of query response covariance matrices is calculated to obtain the target word-first alternative response strategy query response adaptive matching representation matrix.

9. The diabetes intervention intelligent recommendation method based on knowledge graph according to claim 8, characterized in that: Performing local semantic association enhancement on each first candidate response strategy word granularity enhanced semantic coding feature vector in the sequence of the first candidate response strategy word granularity enhanced semantic coding feature vectors to obtain a sequence of first candidate response strategy word granularity local semantic association feature vectors, including: Fine-grained decoupling of the first candidate response strategy word granularity enhanced semantic encoding feature vector is performed to obtain a sequence of first candidate response strategy word granularity local semantic encoding feature vectors; The sequence of the first candidate response strategy word granularity local semantic encoding feature vectors is input into a sequence semantic reinforcement encoder based on the Bi-LSTM model to obtain the first candidate response strategy word granularity local semantic association feature vector.