Knowledge management tool set classification verification and intelligent error correction method and system

Abstract

The invention provides a knowledge management tool set classification verification and intelligent error correction method and system, and relates to the field of intelligent classification verification and error correction in the hydropower industry. The method comprises the following steps: acquiring original knowledge data elements of a knowledge management tool set, synchronously acquiring scene associated metadata of the original knowledge data elements, processing the original knowledge data elements and the scene associated metadata based on a hydropower industry knowledge graph pre-training model, and generating a multi-dimensional scene tag; constructing a water and electricity knowledge association rule base, and inputting the knowledge data set to be verified into a verification model; calling a hydroelectric knowledge enhancement large model and a hydroelectric knowledge error correction case library according to the strategy to be subjected to error correction to obtain knowledge data after error correction; application feedback is carried out, and iterative optimization is carried out; according to the method, the classification accuracy is improved, the complex error recognition rate is increased, meanwhile, the error correction operation is guaranteed by applying feedback in combination with a closed loop of iterative optimization, and the technicality and effectiveness of data response are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent classification verification and error correction in the hydropower industry, and in particular to a knowledge management toolset classification verification and intelligent error correction method and system. Background Technology

[0002] As the hydropower industry transforms towards digitalization and intelligentization, knowledge management toolsets have become the core carrier for integrating diverse and heterogeneous knowledge within the industry, such as equipment operation and maintenance manuals, construction process standards, fault diagnosis cases, and hydrological and meteorological data. Therefore, existing technologies place higher demands on the classification, verification, and intelligent error correction of knowledge management toolsets in the hydropower industry.

[0003] However, existing technologies have the following shortcomings: First, existing data verification technologies (such as CN114997569A) only focus on the general verification of single-type data such as electricity data, and do not design specific solutions for the multi-source heterogeneous knowledge elements of hydropower knowledge management toolsets. Classification rules rely on general levels down to business logic, failing to achieve deep binding between knowledge elements and hydropower business scenarios, resulting in classification verification deviating from actual application needs. Second, existing verification models, such as the six-dimensional monitoring of CN114997569A and the text correction model of CN117609500A, can only perform routine verifications such as format and threshold checks. They do not utilize the industry-specific relationships between hydropower knowledge elements, such as the physical relationship between turbine speed, water level, and flow rate, or the causal relationship between dam seepage pressure anomalies and meteorological rainfall. They cannot identify complex errors such as knowledge logic contradictions and scenario mismatches, resulting in low verification accuracy. Third, existing technologies rely on static preset rules and manual intervention. On the one hand, the verification and correction models cannot adapt to the dynamic updates of hydropower knowledge, such as new unit operation and maintenance knowledge and new industry standards; manual reconfiguration of rules is required, which is inefficient. On the other hand, error correction can only correct formatting errors / typos; logical errors in knowledge require manual modification, and the error correction effect is not verified in connection with downstream application scenarios, which cannot guarantee the effectiveness of knowledge application. Fourth, the existing process only covers the one-way links of data acquisition, classification, verification, and error correction. It has not established a linkage mechanism for knowledge application feedback after error correction and model iteration, which makes it impossible for the verification combined with error correction model to be optimized according to the actual application effect. The problem of correct data but ineffective application has existed for a long time, making it difficult to support the continuous iteration of knowledge management in the hydropower industry. Summary of the Invention

[0004] The main objective of this invention is to provide a knowledge management toolset classification verification and intelligent error correction method and system, which solves the problems of insufficient industry adaptability, limited verification logic, lack of intelligent and dynamic adaptation capabilities, lack of full-link closed loop and low efficiency in the existing technology.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a method for classification verification and intelligent error correction of knowledge management toolsets, comprising the following steps: Obtain the raw knowledge data elements of the knowledge management toolset and simultaneously collect the scenario-related metadata of the raw knowledge data elements; Based on the pre-trained model of the knowledge graph of the hydropower industry, the original knowledge data elements and scene-related metadata are processed to generate multi-dimensional scene tags. Then, exclusive classification rules are generated according to the combination logic of the multi-dimensional scene tags. The original knowledge data elements are classified according to the exclusive classification rules to obtain the knowledge dataset to be verified. A hydropower knowledge association rule base is constructed, defining the relationships between knowledge elements in the form of structured data. The knowledge dataset to be verified is input into the verification model, which includes a basic verification layer and a knowledge enhancement verification layer. The basic verification layer performs routine verification on the format and parameter thresholds of the knowledge elements in the knowledge dataset to be verified. The knowledge enhancement verification layer calls the hydropower knowledge association rule base to perform cross-verification on the knowledge elements and related knowledge elements in the knowledge dataset to be verified, generating a correction strategy containing scenario-based error location information. According to the error correction strategy, the hydropower knowledge enhancement model and the hydropower knowledge error correction case library are invoked to generate multiple candidate error correction schemes. The multiple candidate error correction schemes are then substituted into the hydropower knowledge association rule library for verification. The optimal error correction scheme is selected and the error correction operation is performed to obtain the corrected knowledge data. The application feedback data of the corrected knowledge data is collected, and the verification model and the exclusive classification rules are iteratively optimized based on the application feedback data.

[0006] In the preferred embodiment, the process of processing the original knowledge data elements and scene-related metadata based on the hydropower industry knowledge graph pre-training model to generate multi-dimensional scene tags includes: By acquiring publicly available standards and enterprise internal knowledge base data in the hydropower industry, and fine-tuning the initial pre-trained model, the hydropower industry knowledge graph pre-trained model is obtained. The original knowledge data elements and scene-related metadata are input into the hydropower industry knowledge graph pre-training model. The hydropower industry knowledge graph pre-training model extracts the features of the original knowledge data elements and the key information of the scene-related metadata, automatically labels the multi-dimensional scene tags, and generates a "tag-knowledge element" mapping table.

[0007] In the preferred embodiment, the construction of the hydropower knowledge association rule base includes: Based on the hydropower industry knowledge graph, physical relationships and business causal relationships between different knowledge elements are extracted. The physical relationships include the relationship between turbine speed parameters and basin water level-flow data, and the business causal relationships include the causal relationship between dam seepage pressure monitoring anomalies and meteorological rainfall data and dam displacement monitoring data. The physical relationships and business causal relationships are transformed into structured association logic formulas and causal rules, which are then integrated to form the hydropower knowledge association rule base.

[0008] In the preferred embodiment, the knowledge enhancement verification layer of the verification model calls the hydropower knowledge association rule base to perform cross-verification of knowledge elements and associated knowledge elements in the knowledge dataset to be verified, including: The knowledge enhancement verification layer matches the corresponding associated knowledge elements and association rules from the hydropower knowledge association rule base based on the multi-dimensional scene labels of the knowledge elements in the knowledge dataset to be verified. Based on the association rules, the logical consistency of the knowledge elements in the knowledge dataset to be verified with the matched associated knowledge elements is checked. If there is a logical contradiction, scenario-based error location information containing the contradiction point, the source of the associated data, and the error type is generated.

[0009] In the preferred embodiment, the step of generating multiple candidate error correction schemes by invoking the hydropower knowledge enhancement model and the hydropower knowledge error correction case library according to the error correction strategy to be corrected includes: The hydropower knowledge enhancement model is a model obtained by fine-tuning the open-source model after enhancing hydropower knowledge. The hydropower knowledge error correction case library stores historical correct cases and industry standard solutions. The scenario-based error location information in the error correction strategy is input into the hydropower knowledge enhancement model. The hydropower knowledge enhancement model combines the data in the hydropower knowledge error correction case library to generate multiple candidate error correction schemes for the error. The candidate error correction schemes include error correction content and related supporting cases.

[0010] In the preferred embodiment, the iterative optimization of the verification model and the specific classification rules based on the application feedback data includes: The error types corresponding to the corrected knowledge data of "invalid application" are statistically analyzed, and the error types include scene label errors and missing association rules; Based on the error type, adjust the scene label annotation weight and association rule matching priority of the knowledge enhancement verification layer in the verification model, and update the multi-dimensional scene label combination logic in the exclusive classification rule.

[0011] In a preferred embodiment, a knowledge management toolset classification verification and intelligent error correction system includes: The data acquisition module is used to acquire the original knowledge data elements of the knowledge management toolset and simultaneously collect the scenario-related metadata of the original knowledge data elements; The classification processing module is used to process the original knowledge data elements and scene-related metadata based on the pre-trained model of the hydropower industry knowledge graph, generate multi-dimensional scene labels, generate exclusive classification rules according to the combination logic of the multi-dimensional scene labels, and perform classification operations on the original knowledge data elements according to the exclusive classification rules to obtain the knowledge dataset to be verified. The verification module is used to construct a hydropower knowledge association rule base, which defines the relationship between knowledge elements in the form of structured data. The knowledge dataset to be verified is input into the verification model, which includes a basic verification layer and a knowledge enhancement verification layer. The basic verification layer performs routine verification on the format and parameter thresholds of the knowledge elements in the knowledge dataset to be verified. The knowledge enhancement verification layer calls the hydropower knowledge association rule base to perform cross-verification on the knowledge elements and related knowledge elements in the knowledge dataset to be verified, and generates a correction strategy containing scenario-based error location information. The error correction module is used to call the hydropower knowledge enhancement big model and the hydropower knowledge error correction case library according to the error correction strategy to be corrected, generate multiple candidate error correction schemes, substitute the multiple candidate error correction schemes into the hydropower knowledge association rule library for verification, select the optimal error correction scheme and perform the error correction operation to obtain the corrected knowledge data. The iterative optimization module is used to collect application feedback data of the corrected knowledge data and to iteratively optimize the verification model and the specific classification rules based on the application feedback data.

[0012] In a preferred embodiment, the classification processing module includes: The model training unit is used to acquire publicly available standards and enterprise internal knowledge base data in the hydropower industry, fine-tune the initial pre-trained model, and obtain the hydropower industry knowledge graph pre-trained model. The tag generation unit is used to input the original knowledge data elements and scene-related metadata into the hydropower industry knowledge graph pre-training model. The hydropower industry knowledge graph pre-training model extracts the features of the original knowledge data elements and the key information of the scene-related metadata, automatically labels the multi-dimensional scene tags, and generates a "tag-knowledge element" mapping table. The classification rule generation and execution unit is used to generate exclusive classification rules based on the combination logic of the multi-dimensional scene labels, and to perform classification operations on the original knowledge data elements according to the exclusive classification rules to obtain the knowledge dataset to be verified.

[0013] In a preferred embodiment, the verification module includes: The association rule base construction unit is used to extract physical associations and business causal relationships between different knowledge elements based on the hydropower industry knowledge graph. The physical associations include the association between turbine speed parameters and basin water level-flow data. The business causal relationships include the causal relationship between dam seepage pressure monitoring anomalies and meteorological rainfall data and dam displacement monitoring data. The physical associations and business causal relationships are transformed into structured association logic formulas and causal rules, and integrated to form the hydropower knowledge association rule base. The basic verification unit is used to perform routine verification of the format and parameter thresholds of knowledge elements in the knowledge dataset to be verified through the basic verification layer of the verification model. The enhanced verification unit is used to match corresponding related knowledge elements and association rules from the hydropower knowledge association rule base based on the multi-dimensional scene tags of knowledge elements in the knowledge dataset to be verified through the knowledge enhancement verification layer of the verification model. Based on the association rules, it performs logical consistency verification on the knowledge elements in the knowledge dataset to be verified and the matched related knowledge elements. If there is a logical contradiction, it generates scenario-based error location information containing the contradiction point, the source of related data and the error type, and forms an error correction strategy.

[0014] In the preferred embodiment, the iterative optimization module includes: The feedback data acquisition unit is used to collect application feedback data of the corrected knowledge data in downstream application scenarios. The application feedback data includes knowledge element ID, application scenario tag and effect score. The error statistics unit is used to count the error types corresponding to the corrected knowledge data that is "invalid application". The error types include scene label errors and missing association rules. The optimization unit is used to adjust the scene label annotation weight and association rule matching priority of the knowledge enhancement verification layer in the verification model according to the error type, and at the same time update the multi-dimensional scene label combination logic in the exclusive classification rule.

[0015] This invention provides a method and system for classification verification and intelligent error correction of knowledge management toolsets. It acquires raw knowledge data elements from the knowledge management toolset and simultaneously collects scenario-related metadata for these raw knowledge data elements. Based on a pre-trained model of a hydropower industry knowledge graph, it processes the raw knowledge data elements and scenario-related metadata to generate multi-dimensional scenario labels. A hydropower knowledge association rule library is constructed, containing structured data of knowledge elements, associated knowledge elements, and associated logical formulas / causal rules. The knowledge dataset to be verified is input into the verification model. According to the error correction strategy, a large-scale hydropower knowledge enhancement model and a hydropower knowledge error correction case library are invoked to obtain corrected knowledge data. Application feedback is then provided for iterative optimization. This achieves deep binding between knowledge elements and hydropower business scenarios, improving classification accuracy and complex error recognition rates. It realizes end-to-end intelligent error correction of hydropower knowledge errors, reduces manual intervention costs, and improves data accuracy. Simultaneously, the closed-loop guarantee of application feedback combined with iterative optimization ensures error correction operations, avoiding the problem of correct data but ineffective application, and improving the technicality and effectiveness of data response. Attached Figure Description

[0016] The present invention will be further described below with reference to the accompanying drawings and embodiments: Figure 1 This is a flowchart of the classification verification and intelligent error correction method of the present invention; Figure 2 This is a structural block diagram of the classification verification and intelligent error correction system of the present invention. Detailed Implementation

[0017] Example 1 like Figure 1-2 As shown, a knowledge management toolset classification verification and intelligent error correction method includes the following steps: S1: Obtain the original knowledge data elements of the knowledge management toolset and synchronously collect the scenario-related metadata of the original knowledge data elements. The original knowledge data elements include equipment operation and maintenance manual fragments, construction process standards, fault diagnosis cases and hydrological and meteorological related data. The scenario-related metadata includes the associated equipment number, watershed area, hydrological station and update time information.

[0018] S2: Based on the pre-trained model of the knowledge graph of the hydropower industry, the original knowledge data elements and scene-related metadata are processed to generate multi-dimensional scene tags. The multi-dimensional scene tags include knowledge type tags, business scene tags, related entity tags and time-effect tags. Then, according to the combination logic of the multi-dimensional scene tags, exclusive classification rules are generated. The original knowledge data elements are classified according to the exclusive classification rules to obtain the knowledge dataset to be verified.

[0019] S3: Construct a hydropower knowledge association rule base. The hydropower knowledge association rule base contains structured data of knowledge elements, related knowledge elements, and related logical formulas / causal rules. Input the knowledge dataset to be verified into the verification model. The verification model includes a basic verification layer and a knowledge enhancement verification layer. The basic verification layer performs routine verification on the format and parameter thresholds of the knowledge elements in the knowledge dataset to be verified. The knowledge enhancement verification layer calls the hydropower knowledge association rule base to perform cross-verification between the knowledge elements in the knowledge dataset to be verified and the related knowledge elements, and generates a correction strategy containing scenario-based error location information.

[0020] S4: Based on the error correction strategy, call the hydropower knowledge enhancement model and the hydropower knowledge error correction case library to generate multiple candidate error correction schemes. Substitute the multiple candidate error correction schemes into the hydropower knowledge association rule library for verification, select the optimal error correction scheme and execute the error correction operation to obtain the corrected knowledge data.

[0021] S5: Collect application feedback data of the corrected knowledge data in downstream application scenarios. The application feedback data includes knowledge element ID, application scenario tag and effect score. Iterate and optimize the verification model and exclusive classification rules based on the application feedback data.

[0022] In this embodiment, the original knowledge data elements of the knowledge management toolset are acquired, and the scene-related metadata of the original knowledge data elements is collected simultaneously. Based on the pre-trained model of the hydropower industry knowledge graph, the original knowledge data elements and scene-related metadata are processed to generate multi-dimensional scene labels. A hydropower knowledge association rule library is constructed, which contains structured data of knowledge elements, related knowledge elements, and related logical formulas / causal rules. The knowledge dataset to be verified is input into the verification model. According to the error correction strategy, the hydropower knowledge enhancement model and the hydropower knowledge error correction case library are called to obtain the corrected knowledge data. Application feedback is then provided for iterative optimization. This achieves deep binding between knowledge elements and hydropower business scenarios, improves classification accuracy, improves the recognition rate of complex errors, and realizes end-to-end intelligent error correction of hydropower knowledge errors. This reduces the cost of manual intervention and improves data accuracy. At the same time, the closed-loop guarantee of application feedback combined with iterative optimization ensures the error correction operation, avoids the problem of correct data but ineffective application, and improves the technicality and effectiveness of data response.

[0023] In the preferred scheme, based on the pre-trained model of the hydropower industry knowledge graph, the original knowledge data elements and scene-related metadata are processed to generate multi-dimensional scene tags, including: By acquiring publicly available standards and internal knowledge base data from the hydropower industry, and fine-tuning the initial pre-trained model, a pre-trained knowledge graph model for the hydropower industry is obtained.

[0024] The original knowledge data elements and scene-related metadata are input into the hydropower industry knowledge graph pre-training model. The hydropower industry knowledge graph pre-training model extracts the features of the original knowledge data elements and the key information of the scene-related metadata, automatically labels multi-dimensional scene tags, and generates a "tag-knowledge element" mapping table.

[0025] In this embodiment, the initial pre-trained model is fine-tuned based on the open standards of the hydropower industry and the enterprise knowledge base to obtain a hydropower knowledge graph pre-trained model. Multi-dimensional scene labels are automatically labeled and a "label-knowledge element" mapping table is generated. The automated labeling of multi-dimensional scene labels is achieved through the industry-specific pre-trained model, which improves the labeling efficiency, enhances the matching of label dimensions with the needs of hydropower knowledge management, improves data accuracy, and solves the problem of scene disconnect in the labeling of general models.

[0026] In the preferred scheme, the construction of the hydropower knowledge association rule base includes: Based on the hydropower industry knowledge graph, physical relationships and business causal relationships between different knowledge elements are extracted. Physical relationships include the relationship between turbine speed parameters and water level-flow data in the basin. Business causal relationships include the causal relationship between dam seepage pressure monitoring anomalies and meteorological rainfall data and dam displacement monitoring data.

[0027] Physical relationships and business causal relationships are transformed into structured relational logic formulas and causal rules, which are then integrated to form a hydropower knowledge relational rule base.

[0028] In this embodiment, physical relationships and business causal relationships are extracted based on the hydropower knowledge graph and transformed into structured association logic formulas / causal rules. A hydropower knowledge association rule library is constructed, which enables the verification model to identify hydropower-specific knowledge logic errors, thereby improving the error identification capability. The structured design of the association rule library provides a unified calling interface, which improves the consistency and scalability of cross-validation.

[0029] In the preferred scheme, the knowledge enhancement verification layer of the verification model calls the hydropower knowledge association rule base to perform cross-validation between knowledge elements and associated knowledge elements in the knowledge dataset to be verified, including: The knowledge enhancement verification layer matches the corresponding associated knowledge elements and association rules from the hydropower knowledge association rule base based on the multi-dimensional scene labels of the knowledge elements in the knowledge dataset to be verified.

[0030] Based on association rules, logical consistency checks are performed between knowledge elements in the knowledge dataset to be verified and the matched associated knowledge elements. If logical contradictions exist, scenario-based error location information containing contradiction points, associated data sources, and error types is generated.

[0031] In this embodiment, scenario-based error location information is used, such as errors in turbine efficiency parameters: if the water inflow data of the same basin shows that the efficiency parameter is higher than the design maximum during the dry season, it is recommended to check the source of power generation statistics, etc., which improves the accuracy of error root cause location, shortens error troubleshooting time, and improves work efficiency; logical consistency verification based on hydropower-specific association rules is adopted to avoid the limitation of existing technologies that only verify a single element, improve the accuracy of complex error identification, and improve the knowledge accuracy of hydropower knowledge management tools.

[0032] In the preferred solution, based on the error correction strategy, the hydropower knowledge enhancement model and the hydropower knowledge error correction case library are invoked to generate multiple candidate error correction solutions, including: The hydropower knowledge enhancement model is a model obtained by fine-tuning the open-source model after enhancing hydropower knowledge. The hydropower knowledge error correction case library stores historical correct cases and industry standard solutions. The scenario-based error location information in the error correction strategy is input into the hydropower knowledge enhancement model. The hydropower knowledge enhancement model combines the data in the hydropower knowledge error correction case library to generate multiple candidate error correction solutions for the error. The candidate error correction solutions include error correction content and related supporting cases.

[0033] In this embodiment, multiple candidate solutions provide maintenance personnel with flexible choices, and each solution comes with supporting case studies, such as historical solutions to similar errors and industry standard references. This improves the efficiency of error correction decisions. The application of the hydropower knowledge enhancement model enhances industry adaptability and can generate accurate corrections for logical errors in hydropower knowledge, such as missing key steps in the maintenance process and contradictions between fault causes and related data. This solves the problem that the corrections of general text error correction models are out of touch with industry realities.

[0034] In the preferred solution, iterative optimization of the validation model and specific classification rules based on application feedback data includes: The error types corresponding to the corrected knowledge data for "invalid application" are statistically analyzed. The error types include scene label errors and missing association rules.

[0035] Based on the error type, adjust the scene label annotation weights and association rule matching priorities in the knowledge-enhanced verification layer of the verification model, and update the multi-dimensional scene label combination logic in the exclusive classification rules.

[0036] In this embodiment, a quantitative closed loop for application feedback model optimization is established. By statistically analyzing the types of invalid application errors, the shortcomings in the verification and classification process are accurately identified, such as unreasonable scene label weights and missing association rules, making model optimization more targeted. The classification rules and verification model parameters are dynamically updated, solving the problem that existing static models cannot adapt to knowledge updates, shortening the model's adaptation cycle to new hydropower knowledge, reducing computational load, improving work efficiency, and ensuring the long-term effectiveness of the knowledge management toolset.

[0037] Example 2 To further illustrate with reference to Embodiment 1, a knowledge management toolset classification verification and intelligent error correction system includes: The data acquisition module is used to acquire the original knowledge data elements of the knowledge management toolset and synchronously collect the scenario-related metadata of the original knowledge data elements. The original knowledge data elements include equipment operation and maintenance manual fragments, construction process standards, fault diagnosis cases and hydrological and meteorological related data. The scenario-related metadata includes the associated equipment number, watershed area, hydrological station and update time information.

[0038] The classification processing module is used to process the original knowledge data elements and scene-related metadata based on the pre-trained model of the hydropower industry knowledge graph, and generate multi-dimensional scene tags. The multi-dimensional scene tags include knowledge type tags, business scenario tags, related entity tags and time-effect tags. Then, based on the combination logic of the multi-dimensional scene tags, exclusive classification rules are generated, and the original knowledge data elements are classified according to the exclusive classification rules to obtain the knowledge dataset to be verified.

[0039] The verification module is used to construct a hydropower knowledge association rule base. The hydropower knowledge association rule base contains structured data of knowledge elements, related knowledge elements, and related logical formulas / causal rules. The knowledge dataset to be verified is input into the verification model, which includes a basic verification layer and a knowledge enhancement verification layer. The basic verification layer performs routine verification on the format and parameter thresholds of the knowledge elements in the knowledge dataset to be verified. The knowledge enhancement verification layer calls the hydropower knowledge association rule base to perform cross-verification between the knowledge elements in the knowledge dataset to be verified and the related knowledge elements, and generates a correction strategy containing scenario-based error location information.

[0040] The error correction module is used to generate multiple candidate error correction schemes by calling the hydropower knowledge enhancement big model and the hydropower knowledge error correction case library according to the error correction strategy to be corrected. The multiple candidate error correction schemes are then substituted into the hydropower knowledge association rule library for verification. The optimal error correction scheme is selected and the error correction operation is performed to obtain the corrected knowledge data.

[0041] The iterative optimization module is used to collect application feedback data of the corrected knowledge data in downstream application scenarios. The application feedback data includes knowledge element ID, application scenario tag and effect score. The verification model and exclusive classification rules are iteratively optimized based on the application feedback data.

[0042] In the preferred scheme, the classification processing module includes: The model training unit is used to acquire publicly available standards and internal knowledge base data in the hydropower industry, fine-tune the initial pre-trained model, and obtain a pre-trained knowledge graph model for the hydropower industry.

[0043] The tag generation unit is used to input the original knowledge data elements and scene-related metadata into the hydropower industry knowledge graph pre-training model. The hydropower industry knowledge graph pre-training model extracts the features of the original knowledge data elements and the key information of the scene-related metadata, automatically labels multi-dimensional scene tags, and generates a "tag-knowledge element" mapping table.

[0044] The classification rule generation and execution unit is used to generate exclusive classification rules based on the combination logic of multi-dimensional scene labels, and to perform classification operations on the original knowledge data elements according to the exclusive classification rules to obtain the knowledge dataset to be verified.

[0045] In the preferred embodiment, the verification module includes: The association rule base construction unit is used to extract physical relationships and business causal relationships between different knowledge elements based on the hydropower industry knowledge graph. Physical relationships include the relationship between turbine speed parameters and water level-flow data in the basin. Business causal relationships include the causal relationship between dam seepage pressure monitoring anomalies and meteorological rainfall data and dam displacement monitoring data. The physical relationships and business causal relationships are transformed into structured association logic formulas and causal rules, and integrated to form the hydropower knowledge association rule base.

[0046] The basic verification unit is used to perform routine verification of the format and parameter thresholds of knowledge elements in the knowledge dataset to be verified through the basic verification layer of the verification model.

[0047] The enhanced verification unit is used to match the corresponding related knowledge elements and association rules from the hydropower knowledge association rule base based on the multi-dimensional scene labels of knowledge elements in the knowledge dataset to be verified by the knowledge enhancement verification layer of the verification model. Based on the association rules, the logical consistency verification between the knowledge elements in the knowledge dataset to be verified and the matched related knowledge elements is performed. If there is a logical contradiction, scenario-based error location information containing the contradiction point, the source of the related data and the error type is generated to form an error correction strategy.

[0048] In the preferred scheme, the iterative optimization module includes: The feedback data collection unit is used to collect application feedback data of the corrected knowledge data in downstream application scenarios. The application feedback data includes knowledge element ID, application scenario tag and effect score.

[0049] The error statistics unit is used to count the error types corresponding to the corrected knowledge data that is "invalid application". Error types include scene label errors and missing association rules.

[0050] The optimization unit is used to adjust the scene label annotation weights and association rule matching priorities of the knowledge-enhanced verification layer in the verification model according to the error type, and at the same time update the multi-dimensional scene label combination logic in the exclusive classification rules.

[0051] This embodiment provides the working process, working details, and technical effects of a knowledge management toolset classification verification and intelligent error correction method, which can be referred to in Embodiment 1 and will not be repeated here.

[0052] The above embodiments are merely preferred technical solutions of the present invention and should not be considered as limitations on the present invention. The scope of protection of the present invention should be limited to the technical solutions described in the claims, including equivalent substitutions of the technical features described in the claims. That is, equivalent substitutions and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A method for classification verification and intelligent error correction of knowledge management toolsets, characterized in that, Includes the following steps: Obtain the raw knowledge data elements of the knowledge management toolset and simultaneously collect the scenario-related metadata of the raw knowledge data elements; Based on the pre-trained model of the knowledge graph of the hydropower industry, the original knowledge data elements and scene-related metadata are processed to generate multi-dimensional scene tags. Then, exclusive classification rules are generated according to the combination logic of the multi-dimensional scene tags. The original knowledge data elements are classified according to the exclusive classification rules to obtain the knowledge dataset to be verified. A hydropower knowledge association rule base is constructed, defining the relationships between knowledge elements in the form of structured data. The knowledge dataset to be verified is input into the verification model, which includes a basic verification layer and a knowledge enhancement verification layer. The basic verification layer performs routine verification on the format and parameter thresholds of the knowledge elements in the knowledge dataset to be verified. The knowledge enhancement verification layer calls the hydropower knowledge association rule base to perform cross-verification on the knowledge elements and related knowledge elements in the knowledge dataset to be verified, generating a correction strategy containing scenario-based error location information. According to the error correction strategy, the hydropower knowledge enhancement model and the hydropower knowledge error correction case library are invoked to generate multiple candidate error correction schemes. The multiple candidate error correction schemes are then substituted into the hydropower knowledge association rule library for verification. The optimal error correction scheme is selected and the error correction operation is performed to obtain the corrected knowledge data. The application feedback data of the corrected knowledge data is collected, and the verification model and the exclusive classification rules are iteratively optimized based on the application feedback data.

2. The knowledge management toolset classification verification and intelligent error correction method according to claim 1, characterized in that, The pre-trained model based on the hydropower industry knowledge graph processes the original knowledge data elements and scene-related metadata to generate multi-dimensional scene tags, including: By acquiring publicly available standards and enterprise internal knowledge base data in the hydropower industry, and fine-tuning the initial pre-trained model, the hydropower industry knowledge graph pre-trained model is obtained. The original knowledge data elements and scene-related metadata are input into the hydropower industry knowledge graph pre-training model. The hydropower industry knowledge graph pre-training model extracts the features of the original knowledge data elements and the key information of the scene-related metadata, automatically labels the multi-dimensional scene tags, and generates a "tag-knowledge element" mapping table.

3. The knowledge management toolset classification verification and intelligent error correction method according to claim 1, characterized in that, The construction of the hydropower knowledge association rule base includes: Based on the hydropower industry knowledge graph, physical relationships and business causal relationships between different knowledge elements are extracted. The physical relationships include the relationship between turbine speed parameters and basin water level-flow data, and the business causal relationships include the causal relationship between dam seepage pressure monitoring anomalies and meteorological rainfall data and dam displacement monitoring data. The physical relationships and business causal relationships are transformed into structured association logic formulas and causal rules, which are then integrated to form the hydropower knowledge association rule base.

4. The knowledge management toolset classification verification and intelligent error correction method according to claim 1, characterized in that, The knowledge-enhanced verification layer of the verification model calls the hydropower knowledge association rule base to perform cross-verification of knowledge elements and associated knowledge elements in the knowledge dataset to be verified, including: The knowledge enhancement verification layer matches the corresponding associated knowledge elements and association rules from the hydropower knowledge association rule base based on the multi-dimensional scene labels of the knowledge elements in the knowledge dataset to be verified. Based on the association rules, the logical consistency of the knowledge elements in the knowledge dataset to be verified with the matched associated knowledge elements is checked. If there is a logical contradiction, scenario-based error location information containing the contradiction point, the source of the associated data, and the error type is generated.

5. The knowledge management toolset classification verification and intelligent error correction method according to claim 1, characterized in that, The step of generating multiple candidate error correction schemes by calling the hydropower knowledge enhancement model and the hydropower knowledge error correction case library according to the error correction strategy to be corrected includes: The hydropower knowledge enhancement model is a model obtained by fine-tuning the open-source model after enhancing hydropower knowledge. The hydropower knowledge error correction case library stores historical correct cases and industry standard solutions. The scenario-based error location information in the error correction strategy is input into the hydropower knowledge enhancement model. The hydropower knowledge enhancement model combines the data in the hydropower knowledge error correction case library to generate multiple candidate error correction schemes for the error. The candidate error correction schemes include error correction content and related supporting cases.

6. The knowledge management toolset classification verification and intelligent error correction method according to claim 1, characterized in that, The iterative optimization of the verification model and the specific classification rules based on the application feedback data includes: The error types corresponding to the corrected knowledge data of "invalid application" are statistically analyzed, and the error types include scene label errors and missing association rules; Based on the error type, adjust the scene label annotation weight and association rule matching priority of the knowledge enhancement verification layer in the verification model, and update the multi-dimensional scene label combination logic in the exclusive classification rule.

7. A knowledge management toolset classification verification and intelligent error correction system, characterized in that, include: The data acquisition module is used to acquire the original knowledge data elements of the knowledge management toolset and simultaneously collect the scenario-related metadata of the original knowledge data elements; The classification processing module is used to process the original knowledge data elements and scene-related metadata based on the pre-trained model of the hydropower industry knowledge graph, generate multi-dimensional scene labels, generate exclusive classification rules according to the combination logic of the multi-dimensional scene labels, and perform classification operations on the original knowledge data elements according to the exclusive classification rules to obtain the knowledge dataset to be verified. The verification module is used to construct a hydropower knowledge association rule base, which defines the relationship between knowledge elements in the form of structured data. The knowledge dataset to be verified is input into the verification model, which includes a basic verification layer and a knowledge enhancement verification layer. The basic verification layer performs routine verification on the format and parameter thresholds of the knowledge elements in the knowledge dataset to be verified. The knowledge enhancement verification layer calls the hydropower knowledge association rule base to perform cross-verification on the knowledge elements and related knowledge elements in the knowledge dataset to be verified, and generates a correction strategy containing scenario-based error location information. The error correction module is used to call the hydropower knowledge enhancement big model and the hydropower knowledge error correction case library according to the error correction strategy to be corrected, generate multiple candidate error correction schemes, substitute the multiple candidate error correction schemes into the hydropower knowledge association rule library for verification, select the optimal error correction scheme and perform the error correction operation to obtain the corrected knowledge data. The iterative optimization module is used to collect application feedback data of the corrected knowledge data and to iteratively optimize the verification model and the specific classification rules based on the application feedback data.

8. The knowledge management toolset classification verification and intelligent error correction system according to claim 7, characterized in that, The classification processing module includes: The model training unit is used to acquire publicly available standards and internal knowledge base data in the hydropower industry, fine-tune the initial pre-trained model, and obtain the hydropower industry knowledge graph pre-trained model. The tag generation unit is used to input the original knowledge data elements and scene-related metadata into the hydropower industry knowledge graph pre-training model. The hydropower industry knowledge graph pre-training model extracts the features of the original knowledge data elements and the key information of the scene-related metadata, automatically labels the multi-dimensional scene tags, and generates a "tag-knowledge element" mapping table. The classification rule generation and execution unit is used to generate exclusive classification rules based on the combination logic of the multi-dimensional scene labels, and to perform classification operations on the original knowledge data elements according to the exclusive classification rules to obtain the knowledge dataset to be verified.

9. The knowledge management toolset classification verification and intelligent error correction system according to claim 7, characterized in that, The verification module includes: The association rule base construction unit is used to extract physical associations and business causal relationships between different knowledge elements based on the hydropower industry knowledge graph. The physical associations include the association between turbine speed parameters and basin water level-flow data. The business causal relationships include the causal relationship between dam seepage pressure monitoring anomalies and meteorological rainfall data and dam displacement monitoring data. The physical associations and business causal relationships are transformed into structured association logic formulas and causal rules, and integrated to form the hydropower knowledge association rule base. The basic verification unit is used to perform routine verification of the format and parameter thresholds of knowledge elements in the knowledge dataset to be verified through the basic verification layer of the verification model. The enhanced verification unit is used to match corresponding related knowledge elements and association rules from the hydropower knowledge association rule base based on the multi-dimensional scene tags of knowledge elements in the knowledge dataset to be verified through the knowledge enhancement verification layer of the verification model. Based on the association rules, it performs logical consistency verification on the knowledge elements in the knowledge dataset to be verified and the matched related knowledge elements. If there is a logical contradiction, it generates scenario-based error location information containing the contradiction point, the source of related data and the error type, and forms an error correction strategy.

10. The knowledge management toolset classification verification and intelligent error correction system according to claim 7, characterized in that, The iterative optimization module includes: The feedback data acquisition unit is used to collect application feedback data of the corrected knowledge data in downstream application scenarios. The application feedback data includes knowledge element ID, application scenario tag and effect score. The error statistics unit is used to count the error types corresponding to the corrected knowledge data that is "invalid application". The error types include scene label errors and missing association rules. The optimization unit is used to adjust the scene label annotation weight and association rule matching priority of the knowledge enhancement verification layer in the verification model according to the error type, and at the same time update the multi-dimensional scene label combination logic in the exclusive classification rule.

Images