Digital twinborn-based academic publishing knowledge service intelligent management system and method

Abstract

The invention discloses an academic publication knowledge service intelligent management system and method based on digital twinning, and relates to the field of knowledge service and technology, and the system comprises a data storage module, a dimension weight quantification module, a weight distribution difference analysis module, an effective trigger distribution model construction module and a response priority analysis module. According to the method, the initial weight of each dimension is quantified through six dimensions including multi-dimensional intelligent retrieval, covering knowledge elements, semantic understanding and user portraits, the weight can be dynamically allocated according to user retrieval content features such as whether keywords contain limits or not, new and old attributes of the users and keyword expansion capacity, the weight is updated based on subjective operation contents of the users, and the user experience is improved. The retrieval strategy is deeply matched with the individual requirements of the user, the problem of low retrieval efficiency is greatly reduced, and the operation cost of repeatedly modifying retrieval conditions by the user is reduced.

Description

Technical Field

[0001] This invention relates to the field of knowledge service technology, specifically to an intelligent management system and method for academic publishing knowledge services based on digital twins. Background Technology

[0002] In the current field of academic publishing knowledge services, with the massive growth of digital resources, users' demand for accurate and efficient knowledge retrieval is becoming increasingly urgent. Existing academic publishing knowledge service data management platforms mostly adopt single or fixed-dimensional retrieval methods, lacking the ability to personalize and dynamically adapt to user search behavior. On the one hand, they cannot flexibly adjust the weight of each search dimension based on the characteristics of the user's input search content, such as keyword specificity, demand direction, and knowledge element attributes. This results in a low match between search results and the user's actual needs, often leading to the problem of "low search effectiveness," meaning users need to modify their search criteria multiple times to obtain satisfactory results. On the other hand, the system has not established a correlation model between individual user search behavior and search results, making it difficult to predict the effectiveness of the initial search plan. It also cannot optimize subsequent search strategies based on users' historical subjective operations, such as modifying keywords and filtering search terms. Furthermore, the lack of a scientific division of multi-dimensional search priorities further reduces search efficiency and intelligence. Furthermore, existing systems mostly adopt a unified storage model for managing search data, without tracing and analyzing search behavior on a user-centric basis. This makes it impossible to effectively trace the source of search results and personalize optimization, and it is difficult to meet the management needs of academic publishing knowledge services in the context of digital twin technology for "precision, personalization, and intelligence". Summary of the Invention

[0003] The purpose of this invention is to provide an intelligent management system and method for academic publishing knowledge services based on digital twins, so as to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an intelligent management method for academic publishing knowledge services based on digital twins, the method comprising the following steps: Step S100: Extract retrieval data from the retrieval events of the data management platform used for academic publishing knowledge services using multi-dimensional intelligent retrieval, and store the retrieval data of the retrieval events recorded in each logged-in user as an independent storage unit. Step S200: Based on different retrieval data in the storage unit, quantify the retrieval weights of each dimension of the multi-dimensional intelligent retrieval method and generate the initial weights of different retrieval data in the storage unit. Step S300: Verify the user retrieval effectiveness rate of each retrieval data based on the retrieval event record, extract the weight distribution of each dimension of the retrieval data when the retrieval effectiveness rate is less than the effectiveness rate threshold, and respond to the weight distribution difference. Step S400: Based on the weight distribution difference analysis and the functional relationship of retrieval efficiency, construct an efficiency-triggered allocation model; Step S500: Extract the user subjective operation content in the retrieval events corresponding to retrieval effectiveness rates that are less than the effectiveness rate threshold, and update the initial weights based on the analysis of user subjective operation content when the response effectiveness rate triggers the allocation model. Step S600: Traverse all retrieval events of the same user in the storage unit, output the dimension type for weight update required for each retrieval event, and analyze and generate the multi-dimensional retrieval priority for the corresponding user in the data management platform.

[0005] Furthermore, step S200 includes the following specific details: Multi-dimensional intelligent retrieval includes knowledge element dimension, semantic understanding dimension, resource attribute dimension, user profile dimension, relationship dimension, and multimedia dimension; data platforms that apply multi-dimensional intelligent retrieval methods respond with multi-dimensional data analysis when users enter search content in the search bar; Different intelligent search dimensions assign different weights when analyzing each search result. The specific weights are generated in the following ways: In the semantic understanding dimension, when the input search content contains a single keyword without any limitations, the corresponding weight range [a2, a1] is output; when the input search content contains keywords with a question intent and no limitation on the level or domain, the corresponding weight range [a3, a2) is output; when the input search content contains keywords with directional limitations, the corresponding weight range [a4, a3) is output; a1>a2>a3>a4; In the knowledge element dimension, when all keywords in the input search content are knowledge elements, the corresponding weight interval [b2, b1] is output; when some keywords in the input search content are knowledge elements, the corresponding weight interval [b3, b2) is output; when none of the keywords in the input search content are knowledge elements, the corresponding weight interval [b4, b3) is output; b1>b2>b3>b4; In the user profile dimension, when the keywords in the input search content do not reflect the user's needs and the user is an existing user, the corresponding weight range [c2, c1] is output. Here, an existing user refers to a user whose stored, downloaded, and browsed resource data on the data platform exceeds the data volume threshold. When the keywords in the input search content do not reflect the user's needs and the user is a new user, the corresponding weight range [c3, c2) is output. When the keywords in the input search content reflect a clear need, the corresponding weight range [c4, c3) is output; c1>c2>c3>c4; In terms of relevance, when the number of additional keywords derived from the keywords in the input search content is greater than or equal to a threshold, the corresponding weight interval [d2, d1] is output, typically with a threshold of 2. When the number of additional keywords derived from the keywords in the input search content is greater than zero but less than the threshold, the corresponding weight interval [d3, d2) is output. When the number of additional keywords derived from the keywords in the input search content is zero, the corresponding weight interval [d4, d3) is output. d1>d2>d3>d4. Whether to expand depends on the clarity of the keywords given in the input content. When the keywords are relatively precise, no additional expansion is needed. In the resource attribute dimension, when the number of attribute types corresponding to keywords in the input search content is greater than or equal to the threshold number of types, the corresponding weight interval [e2, e1] is output; when the number of attribute types corresponding to keywords in the input search content is greater than zero but less than the threshold number of types, the corresponding weight interval [e3, e2) is output; when the number of attribute types corresponding to keywords in the input search content is equal to zero, the corresponding weight interval [e4, e3) is output; e1>e2>e3>e4; For the search scope dimension, when the number of keywords in the input search content is greater than or equal to the specified threshold, the corresponding weight interval [f2, f1] is output; when the number of keywords in the input search content is greater than zero but less than the specified threshold, the corresponding weight interval [f3, f2) is output; when the number of keywords in the input search content is zero, the corresponding weight interval [f4, f3) is output; f1>f2>f3>f4; Based on the quantitative indicators under the above different dimensions, the initial weights for the retrieval data of each user's stored records are determined and stored in different dimensions of data analysis.

[0006] Intelligent retrieval quantitative analysis across different dimensions can effectively improve the visualization of data analysis, evaluate the rationality of user-input retrieval data from a quantitative perspective, and provide effective traceability of the results generated by the system's multi-dimensional intelligent retrieval data analysis.

[0007] Furthermore, step S300 includes the following specific steps: Step S310: Determining efficiency includes the following: extracting the subjective actions performed by the user after inputting search data and responding to the search results; subjective actions include modifying or adding to the initial search data and filtering search terms; determining the displayed results content before the user's subjective actions as the first search content, and the displayed results content after the user's subjective actions as the second search content; comparing the similarity between the first search content and the second search content. The similarity result is used as the criterion for judging the effectiveness of user retrieval; when the similarity is less than the similarity threshold, the corresponding retrieval event is marked as the target retrieval event; the retrieval effectiveness of the target retrieval event for the corresponding user is less than the effectiveness threshold. Step S320: Extract the weight values ​​of each dimension in the multi-dimensional intelligent analysis of the first search content response corresponding to the target search event, using the formula: Q={[∑(U1-U0) 2 ] / m} 1 / 2 ; Calculate the weight distribution difference Q under the target retrieval event, where U1 represents the initial weight of each dimension in the target retrieval event, U0 represents the average weight of all dimensions in the target retrieval event record, and m represents the number of dimensions.

[0008] Furthermore, step S400 includes the following specific steps: Step S410: Extract the weight distribution difference Q and search similarity W from the target search event and valid search event of the same user record, and form an array A, A=(Q,W). Valid search events refer to events with search similarity greater than or equal to the similarity threshold and events without recorded user subjective operation. When the user does not perform subjective operation, output search similarity W=1. Step S420: Iterate through all the retrieved data generated by the same user to generate array A, and construct an efficient trigger allocation model w, w=a×Q+ε; where a represents the reference coefficient and ε represents the error term; substitute multiple arrays A of the same user, and use Q as the dependent variable and W as the independent variable into the model to generate an efficient trigger allocation model containing specific reference coefficients and error terms.

[0009] This model can effectively assess whether user satisfaction with the search results provided by the system is affected by the difference in weight distribution; whether a greater difference in weight distribution corresponds to a higher search similarity, i.e., a higher search effectiveness; and the construction of an evaluation model for each user allows for prior analysis of potential search results that do not meet user needs after the system assigns initial weights based on the user's own search content. Furthermore, analysis based on individual user data demonstrates the model's adaptability to individual user profiles, ensuring one model is suitable for one user.

[0010] Furthermore, the analysis of user subjective actions updates the initial weights when the response effectively triggers the allocation model, including the following: Extract keywords from user subjective operations as target keywords. Target keywords are keywords that are different from the initial search data. Match target keywords with quantitative features of different dimensions. When there is a feature range that satisfies the weight range corresponding to a certain dimension, mark the corresponding dimension as the key dimension and store the weight range that satisfies the feature range as the correction weight. The same keyword should be classified into at least one dimension in quantitative feature matching across different dimensions; when the same keyword can be matched to multiple dimensions, the corresponding weights of the multiple dimensions should be updated when the effective triggering of the allocation model is achieved; specifically including the following: The system acquires the search data input by the user, and obtains the initial weights a0, b0, c0, d0, e0, and f0 for each search dimension based on system analysis. It calculates the weight distribution difference corresponding to the initial weights and substitutes them into the effective trigger allocation model for the same user to obtain the estimated search similarity. When the estimated search similarity is less than the similarity threshold, it triggers a response to extract the key dimensions and the corresponding corrected weights for the key dimensions, and then responds to the search again.

[0011] Furthermore, step S600 includes the following steps: The search events that are assigned the same initial weight are classified as a type of main event, and all the search events that require weight updates in the main events are summarized as a target set. Extract the key dimensions for weight updates from the target set, and extract the number of valid retrieval events P1 recorded in the target set for each key dimension. Valid retrieval means that the estimated retrieval similarity obtained by the effective triggering allocation model after weight update is greater than the similarity threshold. Calculate the priority index Y for each key dimension using the formula: Y=P1 / P0; P0 represents the total number of retrieval events for weight updates applied to the key dimensions. Iterate through all key dimensions in the target set to obtain the corresponding priority index Y. Sort the key dimensions in descending order of priority index Y to generate a multi-dimensional search priority under a type of main event. Iterate through all main events with initial weight allocation of all types to generate a multi-dimensional search priority under the corresponding main event.

[0012] Priority analysis is used to provide directional guidance for updating the weights corresponding to multiple search dimensions when the output value of the allocation model triggered by the user's response efficiency is less than the threshold. When there are multiple weight update methods for the same type of subject event, the application of priority can quickly improve the rationality and effectiveness of weight allocation, thereby enhancing the user's preference for multi-dimensional intelligent retrieval in the knowledge service data platform, reducing the impact of subjective biases of different users, and improving the accuracy and intelligence of retrieval.

[0013] The intelligent management system for academic publishing knowledge services based on digital twins includes a data storage module, a dimension weight quantification module, a weight distribution difference analysis module, an efficiency trigger allocation model construction module, and a response priority analysis module. The data storage module is used to extract retrieval data from retrieval events of the data management platform for academic publishing knowledge services using multi-dimensional intelligent retrieval. Each logged-in user is used as an independent storage unit, and the retrieval data of the retrieval event records is stored in the corresponding storage unit. The dimension weight quantization module is used to quantify the retrieval weights of each dimension in the application of multi-dimensional intelligent retrieval methods, and generate storage units to record the initial weights of different retrieval data; The weight distribution difference analysis module is used to extract the weight distribution of each dimension when the retrieval efficiency is less than the efficiency threshold, and respond to the weight distribution difference. The efficient trigger allocation model building module is used to build efficient trigger allocation models; The response priority analysis module is used to update the initial weights based on the analysis of the user's subjective operation content when the response efficiency triggers the allocation model; and it iterates through all retrieval events of the same user in the storage unit, outputs the dimension type that needs to be updated for each retrieval event, and analyzes and generates the corresponding user's multi-dimensional retrieval priority on the data management platform.

[0014] Furthermore, the efficient triggering allocation model building module includes a data group building unit and a model output unit; The data group construction unit is used to extract the weight distribution difference and retrieval similarity between the target retrieval event and the valid retrieval event of the same user record, and form an array; The model output unit iterates through all the search data of the same user to generate an array, and constructs an efficient trigger allocation model.

[0015] Furthermore, the response priority analysis module includes a weight update analysis unit and a priority analysis unit; The weight update analysis unit extracts keywords from user subjective operations as target keywords and matches them with quantitative features of different dimensions. When a feature range that satisfies the weight range corresponding to a certain dimension exists, the corresponding dimension is marked as a key dimension, and the weight range that satisfies the feature range is stored as the corrected weight. The unit also acquires the search data input by the user, analyzes the system to obtain the initial weights of each search dimension, calculates the weight distribution difference corresponding to the initial weights, and substitutes it into the effective trigger allocation model for the same user to obtain the estimated search similarity. When the estimated search similarity is less than the similarity threshold, a response is triggered to extract the key dimension and the corrected weights corresponding to the key dimension for updating, and the search is triggered again. The priority analysis unit is used to classify the retrieval events corresponding to the same initial weight as a type of main event, and to summarize all retrieval events in the main events that require weight updates as a target set; it extracts the number of retrieval events that are valid for updating the retrieval in the target set of each key dimension, and calculates the priority index of each key dimension; it iterates through all key dimensions in the target set to obtain the corresponding priority index Y, and sorts the key dimensions in descending order of priority index to generate a multi-dimensional retrieval priority under a type of main event.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention uses multi-dimensional intelligent retrieval, covering six dimensions including knowledge elements, semantic understanding, and user profiles, to quantify the initial weight of each dimension. It can dynamically allocate weights based on the characteristics of user search content, such as whether keywords contain limitations, user new or old attributes, and keyword expansion capabilities. At the same time, it updates weights based on user subjective operation content, so that the retrieval strategy is deeply matched with the individual needs of users, which greatly reduces the problem of "low retrieval efficiency" and reduces the operation cost of users repeatedly modifying search conditions.

[0017] 2. By constructing an "efficient trigger allocation model" and combining the weight distribution differences and search similarity data of users' historical searches, the effectiveness of the initial search plan can be predicted. When the estimated search similarity is lower than the threshold, the weight update can be actively triggered to avoid invalid search results output, improve the foresight and intelligence of the system response, and the model is built based on individual user data, fully reflecting the subjectivity of user profiles and realizing personalized adaptation of "one person, one model". 3. When weights need to be adjusted, this invention prioritizes higher-priority dimensions for optimization, avoiding the blindness of multi-dimensional adjustments, quickly improving the rationality of weight allocation, and further reducing the impact of user subjective bias on search results. Furthermore, by storing search data with each user as an independent storage unit, combined with the quantitative recording of multi-dimensional weights, it enables full-process traceability of search events, facilitating the analysis of user search behavior patterns and the adaptation of system search strategies. This provides data support for the subsequent management and optimization of academic publishing knowledge services, aligning with the requirements of digital twin technology for "data traceability and optimizable management," and promoting the upgrading of academic publishing knowledge services towards refined management. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the intelligent management method for academic publishing knowledge services based on digital twins, as described in this invention. Detailed Implementation

[0019] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example: Figure 1 As shown, this invention provides an intelligent management method for academic publishing knowledge services based on digital twins, the method comprising the following steps: Step S100: Extract retrieval data from the retrieval events of the data management platform used for academic publishing knowledge services using multi-dimensional intelligent retrieval, and store the retrieval data of the retrieval events recorded in each logged-in user as an independent storage unit. Step S200: Based on different retrieval data in the storage unit, quantify the retrieval weights of each dimension of the multi-dimensional intelligent retrieval method and generate the initial weights of different retrieval data in the storage unit. Step S300: Verify the user retrieval effectiveness rate of each retrieval data based on the retrieval event record, extract the weight distribution of each dimension of the retrieval data when the retrieval effectiveness rate is less than the effectiveness rate threshold, and respond to the weight distribution difference. Step S400: Based on the weight distribution difference analysis and the functional relationship of retrieval efficiency, construct an efficiency-triggered allocation model; Step S500: Extract the user subjective operation content in the retrieval events corresponding to retrieval effectiveness rates that are less than the effectiveness rate threshold, and update the initial weights based on the analysis of user subjective operation content when the response effectiveness rate triggers the allocation model. Step S600: Traverse all retrieval events of the same user in the storage unit, output the dimension type for weight update required for each retrieval event, and analyze and generate the multi-dimensional retrieval priority for the corresponding user in the data management platform.

[0021] Step S200 includes the following specific contents: Multi-dimensional intelligent retrieval includes knowledge element dimension, semantic understanding dimension, resource attribute dimension, user profile dimension, relationship dimension, and multimedia dimension; data platforms that apply multi-dimensional intelligent retrieval methods respond with multi-dimensional data analysis when users enter search content in the search bar; Different intelligent search dimensions assign different weights when analyzing each search result. The specific weights are generated in the following ways: In the semantic understanding dimension, when the input search content contains a single keyword without any limitations, the corresponding weight range [a2, a1] is output; when the input search content contains keywords with a question intent and no limitation on the level or domain, the corresponding weight range [a3, a2) is output; when the input search content contains keywords with directional limitations, the corresponding weight range [a4, a3) is output; a1>a2>a3>a4; In the knowledge element dimension, when all keywords in the input search content are knowledge elements, the corresponding weight interval [b2, b1] is output; when some keywords in the input search content are knowledge elements, the corresponding weight interval [b3, b2) is output; when none of the keywords in the input search content are knowledge elements, the corresponding weight interval [b4, b3) is output; b1>b2>b3>b4; In the user profile dimension, when the keywords in the input search content do not reflect the user's needs and the user is an existing user, the corresponding weight interval [c2, c1] is output. An existing user is defined as a user whose downloaded or viewed resource data on the data platform exceeds a data volume threshold, typically set at 100 data documents. When the keywords in the input search content do not reflect the user's needs and the user is a new user, the corresponding weight interval [c3, c2) is output. When the keywords in the input search content clearly reflect a user's needs, the corresponding weight interval [c4, c3) is output; c1>c2>c3>c4. In terms of relevance, when the number of additional keywords expanded from a keyword in the input search content is greater than or equal to a threshold, the corresponding weight interval [d2, d1] is output. When a search content contains multiple keywords, the evaluation is based on the maximum number of additional keywords that can be expanded. Generally, the threshold is set to 2. When the number of additional keywords expanded from a keyword in the input search content is greater than zero but less than the threshold, the corresponding weight interval [d3, d2) is output. When the number of additional keywords expanded from a keyword in the input search content is zero, the corresponding weight interval [d4, d3) is output. d1>d2>d3>d4. Whether to expand depends on whether the keywords given in the input content are clear. When the keywords are relatively precise, no additional expansion is needed. In the resource attribute dimension, when the number of attribute types corresponding to keywords in the input search content is greater than or equal to the type number threshold, the corresponding weight interval [e2, e1] is output, where the type number threshold is 2; when the number of attribute types corresponding to keywords in the input search content is greater than zero but less than the type number threshold, the corresponding weight interval [e3, e2) is output; when the number of attribute types corresponding to keywords in the input search content is equal to zero, the corresponding weight interval [e4, e3) is output; e1>e2>e3>e4; In terms of the search scope dimension, when the number of keywords in the input search content is greater than or equal to the specified threshold, the threshold value is 2, and the corresponding weight interval [f2, f1] is output. When the number of keywords in the input search content is greater than zero but less than the specified threshold, the corresponding weight interval [f3, f2) is output. When the number of keywords in the input search content is equal to zero, the corresponding weight interval [f4, f3) is output. f1>f2>f3>f4; Based on the quantitative indicators under the above different dimensions, the initial weights for the retrieval data of each user's stored records are determined and stored in different dimensions of data analysis.

[0022] Intelligent retrieval quantitative analysis across different dimensions can effectively improve the visualization of data analysis, evaluate the rationality of user-input retrieval data from a quantitative perspective, and provide effective traceability of the results generated by the system's multi-dimensional intelligent retrieval data analysis.

[0023] Step S300 includes the following specific steps: Step S310: Determining efficiency includes the following: extracting the subjective actions performed by the user after inputting search data and responding to the search results; subjective actions include modifying or adding to the initial search data and filtering search terms; determining the displayed results content before the user's subjective actions as the first search content, and the displayed results content after the user's subjective actions as the second search content; comparing the similarity between the first search content and the second search content. The similarity result is used as the criterion for judging the effectiveness of user retrieval; when the similarity is less than the similarity threshold, the corresponding retrieval event is marked as the target retrieval event; the retrieval effectiveness of the target retrieval event for the corresponding user is less than the effectiveness threshold. Step S320: Extract the weight values ​​of each dimension in the multi-dimensional intelligent analysis of the first search content response corresponding to the target search event, using the formula: Q={[∑(U1-U0) 2 ] / m} 1 / 2 ; Calculate the weight distribution difference Q under the target retrieval event, where U1 represents the initial weight of each dimension in the target retrieval event, U0 represents the average weight of all dimensions in the target retrieval event record, and m represents the number of dimensions.

[0024] Step S400 includes the following specific steps: Step S410: Extract the weight distribution difference Q and search similarity W from the target search event and valid search event of the same user record, and form an array A, A=(Q,W). Valid search events refer to events with search similarity greater than or equal to the similarity threshold and events without recorded user subjective operation. When the user does not perform subjective operation, output search similarity W=1. Step S420: Iterate through all the retrieved data generated by the same user to generate array A, and construct an efficient trigger allocation model w, w=a×Q+ε; where a represents the reference coefficient and ε represents the error term; substitute multiple arrays A of the same user, and use Q as the dependent variable and W as the independent variable into the model to generate an efficient trigger allocation model containing specific reference coefficients and error terms.

[0025] This model can effectively assess whether user satisfaction with the search results provided by the system is affected by the difference in weight distribution; whether a greater difference in weight distribution corresponds to a higher search similarity, i.e., a higher search effectiveness; and the construction of an evaluation model for each user allows for prior analysis of potential search results that do not meet user needs after the system assigns initial weights based on the user's own search content. Furthermore, analysis based on individual user data demonstrates the model's adaptability to individual user profiles, ensuring one model is suitable for one user.

[0026] The initial weights are updated based on the analysis of user subjective actions when the response effectively triggers the allocation model, including the following: Extract keywords from user subjective operations as target keywords. Target keywords are keywords that are different from the initial search data. Match target keywords with quantitative features of different dimensions. When there is a feature range that satisfies the weight range corresponding to a certain dimension, mark the corresponding dimension as the key dimension and store the weight range that satisfies the feature range as the correction weight. For example, if the target keyword is "2022 to 2024", it is clearly a time-related limitation and belongs to the resource attribute dimension; The same keyword should be classified into at least one dimension in quantitative feature matching across different dimensions; when the same keyword can be matched to multiple dimensions, the corresponding weights of the multiple dimensions should be updated when the effective triggering of the allocation model is achieved; specifically including the following: The system acquires the search data input by the user, and obtains the initial weights a0, b0, c0, d0, e0, and f0 for each search dimension based on system analysis. It calculates the weight distribution difference corresponding to the initial weights and substitutes them into the effective trigger allocation model for the same user to obtain the estimated search similarity. When the estimated search similarity is less than the similarity threshold, it triggers a response to extract the key dimensions and the corresponding corrected weights for the key dimensions, and then responds to the search again.

[0027] For example, the initial weights are: knowledge element dimension 20%, semantic understanding dimension 20%, resource attribute dimension 30%, user profile dimension 25%, relationship dimension 30%, and multimedia dimension 20%. If the estimated search similarity obtained by substituting into the model is less than or equal to the similarity threshold, it indicates that the first search content corresponding to the initial search data may not meet the user's search needs. Therefore, it is necessary to analyze and update the key dimensions and corrected weights stored in the historical user data at the same proportion across all dimensions. For example, user profile dimension 25% + corrected weight 20% = updated 45%. The new weight allocation ratio is then used for another search. At this point, the logic of the selection is analyzed and explained. If the key dimensions stored at the same proportion differ, the selection is based on specific priorities according to subsequent analysis. For example, the key dimensions stored at the same proportion may include two types: semantic understanding dimension and association relationship dimension. If the priority of the semantic understanding dimension is greater than that of the association relationship dimension, then the weight update of the semantic understanding dimension is attempted first. In this application, the total weight of the multiple dimensions must be greater than or equal to 70%, with no upper limit of 100%; when the total weight is less than 70%, the weights of each dimension are increased proportionally to make the total greater than or equal to 70%.

[0028] Step S600 includes the following steps: The search events that are assigned the same initial weight are classified as a type of main event, and all the search events that require weight updates in the main events are summarized as a target set. Extract the key dimensions for weight updates from the target set, and extract the number of valid retrieval events P1 recorded in the target set for each key dimension. Valid retrieval means that the estimated retrieval similarity obtained by the effective triggering allocation model after weight update is greater than the similarity threshold. Calculate the priority index Y for each key dimension using the formula: Y=P1 / P0; P0 represents the total number of retrieval events for weight updates applied to the key dimensions. Iterate through all key dimensions in the target set to obtain the corresponding priority index Y. Sort the key dimensions in descending order of priority index Y to generate a multi-dimensional search priority under a type of main event. Iterate through all main events with initial weight allocation of all types to generate a multi-dimensional search priority under the corresponding main event.

[0029] Priority analysis is used to provide directional guidance for updating the weights corresponding to multiple search dimensions when the output value of the allocation model triggered by the user's response efficiency is less than the threshold. When there are multiple weight update methods for the same type of subject event, the application of priority can quickly improve the rationality and effectiveness of weight allocation, thereby enhancing the user's preference for multi-dimensional intelligent retrieval in the knowledge service data platform, reducing the impact of subjective biases of different users, and improving the accuracy and intelligence of retrieval.

[0030] The intelligent management system for academic publishing knowledge services based on digital twins includes a data storage module, a dimension weight quantification module, a weight distribution difference analysis module, an efficiency trigger allocation model construction module, and a response priority analysis module. The data storage module is used to extract retrieval data from retrieval events of the data management platform for academic publishing knowledge services using multi-dimensional intelligent retrieval. Each logged-in user is used as an independent storage unit, and the retrieval data of the retrieval event records is stored in the corresponding storage unit. The dimension weight quantization module is used to quantify the retrieval weights of each dimension in the application of multi-dimensional intelligent retrieval methods, and generate storage units to record the initial weights of different retrieval data; The weight distribution difference analysis module is used to extract the weight distribution of each dimension when the retrieval efficiency is less than the efficiency threshold, and respond to the weight distribution difference. The efficient trigger allocation model building module is used to build efficient trigger allocation models; The response priority analysis module is used to update the initial weights based on the analysis of the user's subjective operation content when the response efficiency triggers the allocation model; and it iterates through all retrieval events of the same user in the storage unit, outputs the dimension type that needs to be updated for each retrieval event, and analyzes and generates the corresponding user's multi-dimensional retrieval priority on the data management platform.

[0031] The efficient trigger allocation model building module includes a data group building unit and a model output unit; The data group construction unit is used to extract the weight distribution difference and retrieval similarity between the target retrieval event and the valid retrieval event of the same user record, and form an array; The model output unit iterates through all the search data of the same user to generate an array, and constructs an efficient trigger allocation model.

[0032] The response priority analysis module includes a weight update analysis unit and a priority analysis unit; The weight update analysis unit extracts keywords from user subjective operations as target keywords and matches them with quantitative features of different dimensions. When a feature range that satisfies the weight range corresponding to a certain dimension exists, the corresponding dimension is marked as a key dimension, and the weight range that satisfies the feature range is stored as the corrected weight. The unit also acquires the search data input by the user, analyzes the system to obtain the initial weights of each search dimension, calculates the weight distribution difference corresponding to the initial weights, and substitutes it into the effective trigger allocation model for the same user to obtain the estimated search similarity. When the estimated search similarity is less than the similarity threshold, a response is triggered to extract the key dimension and the corrected weights corresponding to the key dimension for updating, and the search is triggered again. The priority analysis unit is used to classify the retrieval events corresponding to the same initial weight as a type of main event, and to summarize all retrieval events in the main events that require weight updates as a target set; it extracts the number of retrieval events that are valid for updating the retrieval in the target set of each key dimension, and calculates the priority index of each key dimension; it iterates through all key dimensions in the target set to obtain the corresponding priority index Y, and sorts the key dimensions in descending order of priority index to generate a multi-dimensional retrieval priority under a type of main event.

[0033] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A digital twin-based intelligent management method for academic publishing knowledge services, characterized by: The method includes the following steps: Step S100: Extract retrieval data from the retrieval events of the data management platform used for academic publishing knowledge services using multi-dimensional intelligent retrieval, and store the retrieval data of the retrieval events recorded in each logged-in user as an independent storage unit. Step S200: Based on different retrieval data in the storage unit, quantify the retrieval weights of each dimension of the multi-dimensional intelligent retrieval method and generate the initial weights of different retrieval data in the storage unit. Step S300: Verify the user retrieval effectiveness rate of each retrieval data based on the retrieval event record, extract the weight distribution of each dimension of the retrieval data when the retrieval effectiveness rate is less than the effectiveness rate threshold, and respond to the weight distribution difference. Step S400: Based on the weight distribution difference analysis and the functional relationship of retrieval efficiency, construct an efficiency-triggered allocation model; Step S500: Extract the user subjective operation content in the retrieval events corresponding to retrieval effectiveness rates that are less than the effectiveness rate threshold, and update the initial weights based on the analysis of user subjective operation content when the response effectiveness rate triggers the allocation model. Step S600: Traverse all retrieval events of the same user in the storage unit, output the dimension type for weight update required for each retrieval event, and analyze and generate the multi-dimensional retrieval priority for the corresponding user on the data management platform.

2. The intelligent management method for academic publishing knowledge services based on digital twins according to claim 1, characterized in that: Step S200 includes the following specific contents: The multi-dimensional intelligent retrieval includes knowledge element dimension, semantic understanding dimension, resource attribute dimension, user profile dimension, association relationship dimension, and multimedia dimension; the data platform applying the multi-dimensional intelligent retrieval method responds with multi-dimensional data analysis when the user enters search content in the search bar; Different intelligent search dimensions assign different weights when analyzing each search result. The specific weights are generated in the following ways: In the semantic understanding dimension, when the input search content contains a single keyword without any limitations, the corresponding weight range [a2, a1] is output; when the input search content contains keywords with a question intent and no limitation on the level or domain, the corresponding weight range [a3, a2) is output; when the input search content contains keywords with directional limitations, the corresponding weight range [a4, a3) is output; a1>a2>a3>a4; In the knowledge element dimension, when all keywords in the input search content are knowledge elements, the corresponding weight interval [b2, b1] is output; when some keywords in the input search content are knowledge elements, the corresponding weight interval [b3, b2) is output; when none of the keywords in the input search content are knowledge elements, the corresponding weight interval [b4, b3) is output; b1>b2>b3>b4; In the user profile dimension, when the keywords in the input search content do not reflect the user's needs and the user is an existing user, the corresponding weight range [c2, c1] is output. Here, an existing user refers to a user whose stored, downloaded, and browsed resource data on the data platform exceeds the data volume threshold. When the keywords in the input search content do not reflect the user's needs and the user is a new user, the corresponding weight range [c3, c2) is output. When the keywords in the input search content reflect a clear need, the corresponding weight range [c4, c3) is output; c1>c2>c3>c4; In terms of relevance, when the number of other keywords expanded from the keywords in the input search content is greater than or equal to the threshold, the corresponding weight interval [d2, d1] is output; when the number of other keywords expanded from the keywords in the input search content is greater than zero but less than the threshold, the corresponding weight interval [d3, d2) is output; when the number of other keywords expanded from the keywords in the input search content is zero, the corresponding weight interval [d4, d3) is output; d1>d2>d3>d4; In the resource attribute dimension, when the number of attribute types corresponding to keywords in the input search content is greater than or equal to the threshold number of types, the corresponding weight interval [e2, e1] is output; when the number of attribute types corresponding to keywords in the input search content is greater than zero but less than the threshold number of types, the corresponding weight interval [e3, e2) is output; when the number of attribute types corresponding to keywords in the input search content is equal to zero, the corresponding weight interval [e4, e3) is output; e1>e2>e3>e4; For the search scope dimension, when the number of keywords in the input search content is greater than or equal to the specified threshold, the corresponding weight interval [f2, f1] is output; when the number of keywords in the input search content is greater than zero but less than the specified threshold, the corresponding weight interval [f3, f2) is output; when the number of keywords in the input search content is zero, the corresponding weight interval [f4, f3) is output; f1>f2>f3>f4; Based on the quantitative indicators under the above different dimensions, the initial weights for the retrieval data of each user's stored records are determined and stored in different dimensions of data analysis.

3. The intelligent management method for academic publishing knowledge services based on digital twins according to claim 2, characterized in that: Step S300 includes the following specific steps: Step S310: Determining efficiency includes the following: extracting the subjective operations performed by the user after responding to the search results by inputting search data, the subjective operations including modifying or adding to the initial search data, and filtering search items; The results displayed before the user's subjective action are designated as the first search result, and the results displayed after the user's subjective action are designated as the second search result. The similarity between the first and second search results is then compared. The similarity results are used as the basis for judging the effectiveness of user searches; when the similarity is less than the similarity threshold, the corresponding search event is marked as the target search event. The output shows that the retrieval success rate of the target retrieval event for the corresponding user is less than the success rate threshold. Step S320: Extract the weight values ​​of each dimension in the multi-dimensional intelligent analysis of the first search content response corresponding to the target search event, using the formula: Q={[∑(U1-U0) 2 ] / m} 1 / 2 ; Calculate the weight distribution difference Q under the target retrieval event, where U1 represents the initial weight of each dimension in the target retrieval event, U0 represents the average weight of all dimensions in the target retrieval event record, and m represents the number of dimensions.

4. The intelligent management method for academic publishing knowledge services based on digital twins according to claim 3, characterized in that: Step S400 includes the following specific steps: Step S410: Extract the weight distribution difference Q and search similarity W from the target search event and valid search event of the same user record, and form an array A, A=(Q,W). The valid search event refers to the event with search similarity greater than or equal to the similarity threshold and the event without recorded user subjective operation. When the user does not perform subjective operation, output search similarity W=1. Step S420: Iterate through all the retrieved data generated by the same user to generate array A, and construct an efficient trigger allocation model w, w=a×Q+ε; where a represents the reference coefficient and ε represents the error term; substitute multiple arrays A of the same user, and use Q as the dependent variable and W as the independent variable into the model to generate an efficient trigger allocation model containing specific reference coefficients and error terms.

5. The intelligent management method for academic publishing knowledge services based on digital twins according to claim 4, characterized in that: The analysis of user subjective operations updates the initial weights when the response effectively triggers the allocation model, including the following: Extract keywords from the user's subjective operation content as target keywords, where the target keywords are keywords different from the initial search data; match the target keywords with quantitative features of different dimensions; when there is a feature range that satisfies the weight range corresponding to a certain dimension, mark the corresponding dimension as the key dimension and store the weight range that satisfies the feature range as the correction weight; The same keyword should be classified into at least one dimension in quantitative feature matching across different dimensions; when the same keyword can be matched to multiple dimensions, the corresponding weights of the multiple dimensions should be updated when the effective triggering of the allocation model is achieved; specifically including the following: The system acquires the search data input by the user, and obtains the initial weights a0, b0, c0, d0, e0, and f0 for each search dimension based on system analysis. It calculates the weight distribution difference corresponding to the initial weights and substitutes them into the effective trigger allocation model for the same user to obtain the estimated search similarity. When the estimated search similarity is less than the similarity threshold, it triggers a response to extract the key dimensions and the corresponding corrected weights for the key dimensions, and then responds to the search again.

6. The intelligent management method for academic publishing knowledge services based on digital twins according to claim 4, characterized in that: Step S600 includes the following steps: The search events that are assigned the same initial weight are classified as a type of main event, and all the search events that require weight updates in the main events are summarized as a target set. Extract the key dimensions for weight updates from the target set, and extract the number of valid retrieval events P1 recorded in the target set for each key dimension. Valid retrieval means that the estimated retrieval similarity obtained by the effective trigger allocation model after weight update is greater than the similarity threshold. Calculate the priority index Y for each key dimension using the formula: Y=P1 / P0; P0 represents the total number of retrieval events for weight updates applied to the key dimensions. Iterate through all key dimensions in the target set to obtain the corresponding priority index Y. Sort the key dimensions in descending order of priority index Y to generate a multi-dimensional search priority under a type of main event. Iterate through all main events with initial weight allocation of all types to generate a multi-dimensional search priority under the corresponding main event.

7. An intelligent management system for academic publishing knowledge services based on digital twins, using the intelligent management method for academic publishing knowledge services based on digital twins as described in any one of claims 1-6, characterized in that, The system includes a data storage module, a dimension weight quantification module, a weight distribution difference analysis module, an efficient trigger allocation model construction module, and a response priority analysis module. The data storage module is used to extract retrieval data from retrieval events of the data management platform for academic publishing knowledge services using multi-dimensional intelligent retrieval. Each logged-in user is used as an independent storage unit, and the retrieval data of the retrieval event records is stored in the corresponding storage unit. The dimension weight quantization module is used to quantify the retrieval weights of each dimension when applying the multi-dimensional intelligent retrieval method, and generate storage units to record the initial weights of different retrieval data. The weight distribution difference analysis module is used to extract the weight distribution of each dimension when the retrieval efficiency is less than the efficiency threshold, and respond to the weight distribution difference. The efficient trigger allocation model construction module is used to construct an efficient trigger allocation model; The response priority analysis module is used to update the initial weights based on the analysis of the user's subjective operation content when the response effectiveness triggers the allocation model. It iterates through all retrieval events of the same user in the storage unit, outputs the dimension type for weight updates required for each retrieval event, and analyzes and generates the corresponding multi-dimensional retrieval priority for the user on the data management platform.

8. The intelligent management system for academic publishing knowledge services based on digital twins according to claim 7, characterized in that: The efficient trigger allocation model construction module includes a data group construction unit and a model output unit; The data group construction unit is used to extract the weight distribution difference and retrieval similarity in the target retrieval event and valid retrieval event of the same user record, and form an array; The model output unit iterates through all the search data of the same user to generate an array, and constructs an efficient trigger allocation model.

9. The intelligent management system for academic publishing knowledge services based on digital twins according to claim 7, characterized in that: The response priority analysis module includes a weight update analysis unit and a priority analysis unit; The weight update analysis unit is used to extract keywords from the user's subjective operation content as target keywords, and match the target keywords with quantitative features of different dimensions. When there is a feature range that satisfies the weight range corresponding to a certain dimension, the corresponding dimension is marked as a key dimension and the weight range that satisfies the feature range is stored as the corrected weight. The unit also obtains the search data input by the user, analyzes the system to obtain the initial weight of each search dimension, calculates the weight distribution difference degree corresponding to the initial weight, and substitutes it into the effective trigger allocation model of the same user to obtain the estimated search similarity. When the estimated search similarity is less than the similarity threshold, the unit triggers a response to extract the key dimension and the corrected weight corresponding to the key dimension for update, and then responds to the search again. The priority analysis unit is used to assign the corresponding retrieval events with the same initial weight as a type of main event, and to summarize all retrieval events in the main events that require weight updates as a target set. Extract the number of valid search events recorded in the target set for each key dimension, and calculate the priority index for each key dimension; traverse all key dimensions in the target set to obtain the corresponding priority index Y, and sort the key dimensions in descending order of priority index to generate a multi-dimensional search priority under a type of main event.

Images