A multi-source heterogeneous data knowledge base intelligent retrieval and management method and system

Abstract

The application discloses a multi-source heterogeneous data knowledge base intelligent retrieval and management method and system, relates to the technical field of data processing and knowledge management, and aims to solve the problems of low retrieval efficiency, low knowledge reuse rate and high system maintenance cost in enterprise multi-source heterogeneous data management. The method comprises the following steps: continuously collecting multi-dimensional behavior and qualification data of users in an enterprise, and constructing and maintaining a sub-field experience level file of the users based on the multi-dimensional behavior and qualification data; in response to a retrieval request of a user, obtaining to-be-retrieved knowledge matched with the retrieval request from a to-be-retrieved knowledge set; according to a technical field identifier of the to-be-retrieved knowledge, extracting the sub-field experience level file of the current user in the corresponding technical field, and determining the risk level of the to-be-retrieved knowledge corresponding to the current user; and according to the risk level, dynamically adjusting the presentation mode of the to-be-retrieved knowledge in the retrieval result.

Description

Technical Field

[0001] This application relates to the fields of data processing and knowledge management technology, and in particular to a method and system for intelligent retrieval and management of multi-source heterogeneous data knowledge base. Background Technology

[0002] In multi-domain business scenarios within enterprises, information such as consulting cases, training courses, and business data are often scattered across different systems, resulting in significant differences in data formats, low retrieval efficiency, and difficulties in accessing this information across different platforms. Traditional management methods struggle to achieve unified indexing and accurate matching of this diverse data, leading not only to low knowledge reuse rates within the enterprise but also to increased system maintenance costs. To address these challenges, an intelligent retrieval and management method and system for multi-source heterogeneous data knowledge bases has emerged. This system aims to aggregate, index, and manage various dispersed information assets within an enterprise, thereby improving knowledge reuse rates and reducing maintenance costs.

[0003] However, even with such a system, new challenges arise in practical applications. For example, in knowledge base systems that integrate multi-source heterogeneous data, including video, existing retrieval technologies primarily rely on direct analysis of data content. This includes extracting keywords and object information from videos through speech recognition and image recognition, and then using this as a basis for matching and recommendation. This approach ignores the "contextual attributes" of knowledge itself—the background and intent behind the information. The system cannot effectively distinguish between a live video recording a real malfunction and a simulated demonstration video used for teaching, even if the two highly overlap in content keywords and visual elements.

[0004] This leads to confusion between real problem diagnosis information and simulated, idealized teaching information in the search results. This not only increases the cognitive burden on users in filtering information, but more seriously, when users (especially inexperienced technicians) misapply solutions from teaching scenarios to real, complex production scenarios, they may make incorrect decisions due to ignoring scenario differences, leading to safety risks or economic losses. The root of the problem is that while the system can understand "what" the data is (identifying bearings and keywords), it cannot understand "what scenario" in which this data exists (whether it is real fault diagnosis or simulated teaching training), thus failing to provide users with scenario-specific, safe, and reliable knowledge support. Summary of the Invention

[0005] This application provides an intelligent retrieval and management method and system for multi-source heterogeneous data knowledge bases, aiming to solve the problems of low retrieval efficiency, low knowledge reuse rate, high system maintenance cost, and the inability of existing retrieval technologies to effectively distinguish knowledge context attributes, resulting in confusing retrieval results and increased user decision-making risks in the management of multi-source heterogeneous data in enterprises.

[0006] Firstly, to address the aforementioned technical problems, this invention provides an intelligent retrieval and management method for a multi-source heterogeneous data knowledge base. This method includes: continuously collecting multi-dimensional behavioral and qualification data of users within an enterprise, and constructing and maintaining user-specific experience level profiles based on this data. These profiles characterize users' knowledge comprehension, risk identification, and operational execution capabilities in different technical fields, with each technical field corresponding to an independent experience level profile. In response to a user's retrieval request, the method retrieves retrieval knowledge matching the request from a set of retrieval knowledge. This set contains all knowledge content stored after unified indexing of multi-source heterogeneous data, and each piece of retrieval knowledge includes a corresponding technical field identifier. Based on the technical field identifier of the retrieval knowledge, the method extracts the current user's sub-domain experience level profile in the corresponding technical field. Based on the sub-domain experience level profile and the contextual attribute information of the retrieval knowledge, the method determines the risk level of the retrieval knowledge corresponding to the current user. Based on the risk level, the method dynamically adjusts the presentation of the retrieval knowledge in the retrieval results and generates operational suggestions matching the sub-domain experience level profile. These suggestions include supplementary knowledge recommendations related to the retrieval knowledge and adapted to the current user's capabilities.

[0007] Secondly, this application provides an intelligent retrieval and management system for multi-source heterogeneous data knowledge bases. This system includes: a collection unit, used to continuously collect multi-dimensional behavioral and qualification data of users within an enterprise, and to construct and maintain user-specific experience level profiles based on this data. These profiles characterize users' knowledge comprehension, risk identification, and operational execution capabilities in different technical fields, with each technical field corresponding to an independent experience level profile; a matching unit, used to respond to user retrieval requests by retrieving knowledge matching the retrieval request from a set of knowledge to be retrieved. This set contains all knowledge content stored after unified indexing of multi-source heterogeneous data, and each piece of knowledge to be retrieved includes a corresponding technical field identifier; an extraction unit, used to extract the current user's specific experience level profile in the corresponding technical field based on the technical field identifier of the knowledge to be retrieved, and to determine the risk level of the knowledge to be retrieved corresponding to the current user based on the specific experience level profile and the contextual attribute information of the knowledge to be retrieved; and a generation unit, used to dynamically adjust the presentation of the knowledge to be retrieved in the retrieval results based on the risk level, and to generate operational suggestions matching the specific experience level profile. These operational suggestions include supplementary knowledge recommendations related to the knowledge to be retrieved and adapted to the current user's capabilities.

[0008] This application offers at least the following advantages: The intelligent retrieval and management method for multi-source heterogeneous data knowledge base disclosed in this application continuously collects multi-dimensional behavioral and qualification data of users within an enterprise, and constructs and maintains user-specific experience level profiles based on this data. This accurately characterizes users' knowledge understanding, risk identification, and operational execution capabilities in different technical fields. When responding to user retrieval requests, this method not only retrieves matching knowledge from multi-source heterogeneous data, but more importantly, it intelligently determines the risk level of the knowledge corresponding to the current user based on the technical field identifier and contextual attribute information of the knowledge, combined with the user's specific experience level profile. Finally, it dynamically adjusts the presentation of knowledge in the retrieval results based on this risk level and generates operational suggestions and supplementary knowledge recommendations tailored to the user's capabilities. Attached Figure Description

[0009] Figure 1 This is one of the flowcharts illustrating an intelligent retrieval and management method for multi-source heterogeneous data knowledge bases provided in this application.

[0010] Figure 2 This is one of the flowcharts illustrating an intelligent retrieval and management method for multi-source heterogeneous data knowledge bases provided in this application.

[0011] Figure 3 This is a schematic diagram of the composition of an intelligent retrieval and management system for multi-source heterogeneous data knowledge base provided in this application. Detailed Implementation

[0012] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0013] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0014] In multi-domain business scenarios within enterprises, information such as consulting cases, training courses, and business data are often scattered across different systems, resulting in significant differences in data formats, low retrieval efficiency, and difficulties in accessing this information across different platforms. Traditional management methods struggle to achieve unified indexing and accurate matching of this diverse data, leading not only to low internal knowledge reuse rates but also increased system maintenance costs. Even with such a system, new challenges arise in practical applications. For example, in knowledge base systems integrating multi-source heterogeneous data, including videos, existing retrieval technologies primarily rely on direct analysis of data content, such as extracting keywords and object information from videos through speech recognition and image recognition, and then using this as a basis for matching and recommendation. This method ignores the "contextual attributes" of the knowledge itself, i.e., the background and intent behind the information's creation. The system cannot effectively distinguish between a live video recording a real malfunction and a simulated demonstration video used for teaching, even if the two highly overlap in content keywords and visual elements. This leads to confusion between real problem diagnosis information and simulated, idealized teaching information in the search results. This not only increases the cognitive burden on users in filtering information, but more seriously, when users (especially inexperienced technicians) misapply solutions from teaching scenarios to real, complex production scenarios, they may make incorrect decisions due to ignoring scenario differences, leading to safety risks or economic losses. The root of the problem is that while the system can understand "what" the data is (identifying bearings and keywords), it cannot understand "what scenario" in which this data exists (whether it is real fault diagnosis or simulated teaching training), thus failing to provide users with scenario-specific, safe, and reliable knowledge support.

[0015] In view of the above problems, this application provides an intelligent retrieval and management method for multi-source heterogeneous data knowledge bases. This method continuously collects multi-dimensional behavioral and qualification data of users within an enterprise, and constructs and maintains user-specific experience level profiles based on this data. This profile characterizes users' knowledge comprehension, risk identification, and operational execution capabilities in different technical fields. When responding to a user's retrieval request, this method can retrieve the retrieval knowledge matching the request from the retrieval knowledge set, and extract the current user's specific experience level profile in the corresponding technical field based on the technical field identifier of the retrieval knowledge. Furthermore, by combining the specific experience level profile and the contextual attribute information of the retrieval knowledge, this method can determine the risk level of the current user corresponding to the retrieval knowledge. Finally, the method dynamically adjusts the presentation of the retrieval knowledge in the retrieval results based on the risk level, and generates operational suggestions matching the specific experience level profile. These suggestions include supplementary knowledge recommendations related to the retrieval knowledge and adapted to the current user's capabilities. In this way, this application aims to solve the problem of mismatch between knowledge retrieval results and user capabilities and knowledge contextual attributes in existing technologies, thereby improving the accuracy, security, and practicality of knowledge retrieval.

[0016] The following specific embodiments will provide a detailed introduction and explanation of the intelligent retrieval and management method for multi-source heterogeneous data knowledge bases provided in this application.

[0017] Reference Figure 1 This application provides an intelligent retrieval and management method for multi-source heterogeneous data knowledge bases, which may include the following steps: S101. Continuously collect multi-dimensional behavioral and qualification data of users within the enterprise, and build and maintain user experience level profiles in different fields based on the multi-dimensional behavioral and qualification data.

[0018] Among them, the domain-specific experience level profiles are used to characterize users’ knowledge comprehension, risk identification and operational execution capabilities in different technical fields. Each technical field corresponds to an independent experience level profile.

[0019] The "multi-source heterogeneous data" mentioned in this application refers to data originating from different systems within an enterprise and possessing different formats and structures, such as documents, structured business data, and videos. After unified indexing, this data forms a "knowledge set to be retrieved," where each piece of knowledge includes a "technical field identifier" to indicate its professional domain. A user's "domain-specific experience level profile" is constructed based on the user's multi-dimensional behavioral and qualification data within the enterprise. It quantifies the user's knowledge understanding, risk identification, and operational execution capabilities within a specific technical field, with each technical field having its own independent experience level profile. Furthermore, "contextual attribute information" are pre-annotated attributes in the metadata of each piece of knowledge to be retrieved, used to describe the background and intent behind the knowledge's creation, such as whether it is record-type content from a real business scenario or simulation-type content from a teaching and training scenario. Based on this information, the system can determine the "risk level" of the knowledge for the current user and adjust the "presentation method" of the knowledge accordingly, generating "operational suggestions" to provide personalized, secure, and reliable knowledge services.

[0020] Specifically, the method in this application first involves continuously collecting multi-dimensional behavioral and qualification data of users within the enterprise. This data can be manually entered, for example, collected and input into the system periodically by administrators; or it can be automatically obtained through system integration, such as obtaining user operation records, task completion status, training certification information, etc., from internal business systems and learning platforms via API interfaces. The collected data is used to construct and maintain user experience level profiles categorized by field. For example, independent experience level profiles can be established for each user in different technical fields such as "equipment maintenance," "software development," and "financial management." These profiles can be easily maintained through manual evaluation and regular updates, such as rating employees based on their performance by department managers or experts.

[0021] S102. In response to the user's search request, retrieve the search knowledge that matches the search request from the search knowledge set.

[0022] The knowledge set to be retrieved contains all knowledge content stored after multi-source heterogeneous data has been indexed in a unified manner, and each piece of knowledge to be retrieved contains a corresponding technical field identifier.

[0023] When a user initiates a search request, the system retrieves matching knowledge from the knowledge set to be searched. This knowledge set can be a centralized database containing all heterogeneous data from multiple sources and indexed uniformly. The matching process can be based on keyword matching; for example, if a user enters "pump troubleshooting," the system will return knowledge whose titles or content contain "pump troubleshooting."

[0024] S103. Based on the technical field identifier of the knowledge to be retrieved, extract the current user's sub-field experience level profile in the corresponding technical field. Based on the sub-field experience level profile and the contextual attribute information of the knowledge to be retrieved, determine the risk level of the current user corresponding to the knowledge to be retrieved.

[0025] Subsequently, based on the technical field identifier of the knowledge to be retrieved, the system will extract the current user's sub-domain experience level profile in the corresponding technical field. For example, if the technical field identifier of the knowledge to be retrieved is "equipment repair," the system will extract the user's experience level profile in the "equipment repair" field. This extraction process can be a simple database query operation.

[0026] Next, the system determines the risk level of the knowledge to be retrieved for the current user based on the user's experience level profile and the contextual attribute information of the knowledge to be retrieved. For example, if the user's experience level profile shows insufficient experience in the "equipment maintenance" field, while the contextual attribute information of the knowledge to be retrieved indicates that it is "record-type content generated in a real business scenario," the system may determine that the knowledge poses a high risk to the current user. The risk level can be determined based on a preset rule table, such as cross-matching the user's experience level with the contextual attributes of the knowledge to obtain high, medium, and low risk levels.

[0027] S104. Based on the risk level, dynamically adjust the presentation of the knowledge to be retrieved in the search results, and generate operational suggestions that match the experience level profiles of different sub-domains.

[0028] The operation suggestions include supplementary knowledge recommendations related to the knowledge to be searched and adapted to the current user's capabilities.

[0029] Finally, based on the determined risk level, the system dynamically adjusts the presentation of the knowledge being retrieved in the search results and generates operational suggestions that match the user's experience level profile across different domains. For example, for high-risk knowledge, a prominent red warning icon can be added to the search results; for medium-risk knowledge, auxiliary prompts can be added to remind users to use it with caution; and for low-risk knowledge, it can be presented in the default manner. The generation of operational suggestions can be based on predefined templates. For instance, when a user lacks experience and the knowledge risk is high, the system can recommend relevant basic training courses or expert contact information.

[0030] The overall working principle of this application lies in establishing user experience level profiles in different technical fields and combining this with the contextual attributes of the knowledge itself to perform personalized risk assessment and presentation of search results. When a user initiates a search request, the system not only matches the content, but more importantly, it intelligently extracts the user's experience level in that field based on the technical field identifier of the knowledge to be searched. Simultaneously, the system identifies the contextual attributes of the knowledge to be searched, such as whether it is a real-world case or a teaching simulation. By comprehensively judging the user's experience level and the contextual attributes of the knowledge, the system can determine the potential risk level of the knowledge for the current user. For example, when an inexperienced user reviews a complex troubleshooting record in a real business scenario, the system will identify a high risk. Based on this risk assessment, the system dynamically adjusts the way knowledge is presented in the search results; for example, it prominently warns high-risk knowledge and provides operational suggestions matching the user's capabilities, such as recommending more basic supplementary knowledge or expert support. This mechanism ensures that the knowledge acquired by the user is not only relevant in content but also adapted to the user's actual capabilities in terms of context and difficulty, thereby effectively avoiding potential risks caused by the misuse of knowledge.

[0031] In some embodiments of this application, reference is made to Figure 2 The aforementioned continuous collection of multi-dimensional behavioral and qualification data of users within an enterprise can specifically include the following steps: S201. Collect user behavior data in the enterprise's multi-source heterogeneous data knowledge base system.

[0032] The operational behavior data includes search records, document browsing time, and video viewing records, with each piece of operational behavior data associated with a corresponding technical field identifier.

[0033] S202. Collect user task execution data in the enterprise business system.

[0034] The task execution data includes the types and quantities of maintenance tasks completed by the user, and each piece of task execution data is associated with a corresponding technical field identifier.

[0035] S203. Collect user training and certification data and evaluation feedback data.

[0036] The training and certification data includes the completion status of internal training courses and external professional certification information. The evaluation and feedback data includes structured evaluations from experts and collaborative feedback from peers. Each piece of training and certification data and evaluation and feedback data is associated with a corresponding technical field identifier.

[0037] S204. Combine operational behavior data, task execution data, training and certification data, and evaluation feedback data into a multi-dimensional set of behavior and qualification data.

[0038] Specifically, operational behavior data refers to the interaction traces of users within an enterprise's multi-source heterogeneous data knowledge base system. Examples include which keywords a user searched, which documents they consulted, and which instructional videos they watched. This data directly reflects a user's interest, level of engagement, and initial understanding of knowledge in a specific technical field. Each piece of operational behavior data is assigned a corresponding technical field identifier to facilitate subsequent classification and analysis.

[0039] Furthermore, task execution data refers to the specific tasks completed by users in actual business operations within the enterprise's business system, such as the number of equipment repair tasks a user successfully completed or the types of faults handled. This type of data directly reflects a user's practical operational capabilities and accumulated experience in a specific technical field. Similarly, each piece of task execution data is also associated with a corresponding technical field identifier.

[0040] Furthermore, training and certification data, along with evaluation and feedback data, reflect users' professional competence from a more formal and authoritative perspective. Training and certification data includes professional skills training courses completed within the company and certifications obtained from external professional organizations, serving as proof of users' systematic learning and professional skill improvement. Evaluation and feedback data includes structured evaluations of user performance from experts and feedback from colleagues during collaboration, providing an objective assessment of user capabilities and collaborative performance. Each piece of training and certification data, as well as evaluation and feedback data, is also associated with a corresponding technical field identifier.

[0041] Ultimately, the aforementioned operational behavior data, task execution data, training and certification data, and evaluation feedback data are integrated to form the multi-dimensional behavior and qualification data, providing a rich data foundation for comprehensively assessing user capabilities.

[0042] Through the aforementioned technical solution, this application achieves comprehensive coverage and refined collection of user behavior and qualification data. This integration of multi-source heterogeneous data makes the constructed user domain-specific experience level profiles more accurate and comprehensive, more realistically reflecting users' knowledge comprehension, risk identification, and operational execution capabilities in different technical fields. Therefore, in subsequent knowledge retrieval and management processes, the system can provide more accurate risk assessments and personalized operational suggestions based on more reliable user profiles, significantly improving the intelligence level of knowledge services and user experience.

[0043] In some of the embodiments described above in this application, a method for constructing and maintaining user-specific experience level profiles based on multi-dimensional behavioral and qualification data is proposed. However, in actual implementation, effectively transforming multi-source, heterogeneous behavioral and qualification data containing non-numerical data into unified, quantifiable experience assessments, and ensuring that the experience level profiles can be dynamically adjusted in real time as user behavior updates, is a technical problem that needs to be solved. If this problem is not solved, the constructed experience level profiles may not accurately reflect the user's latest capability status, thereby affecting the accuracy of subsequent risk level judgments and operational recommendations.

[0044] In this regard, this application further proposes the following steps for constructing and maintaining user-specific experience level profiles based on multi-dimensional behavioral and qualification data: Perform a preset quantization transformation on the non-numerical data in the multi-dimensional behavior and qualification data to obtain the corresponding quantized values; Based on a preset set of weighting coefficients, all quantitative values ​​in the same technical field are weighted to obtain the user's experience score in the corresponding technical field. Based on the experience rating, the user's experience level in the corresponding technical field is divided into multiple preset level categories; When new user operation behavior data, task execution data, training certification data, and evaluation feedback data are obtained, the quantitative conversion processing, weighted summation processing, and level classification processing are repeatedly executed to update the user's experience score and corresponding experience level category in the corresponding technical field in real time.

[0045] Specifically, the multi-dimensional behavioral and qualification data may include various forms, some of which are non-numerical. For example, the "internal training course completion status" in user training certification data may exist in text form such as "completed" or "incomplete," or the "expert structured evaluation" in evaluation feedback data may contain descriptive text. To enable a unified quantitative assessment of this heterogeneous data, a pre-defined quantitative transformation process needs to be performed on this non-numerical data. For example, "completed" can be converted to the value 1, and "incomplete" to the value 0. For descriptive evaluations, keywords can be extracted using natural language processing technology and mapped to a pre-defined scoring range, or human experts can perform scoring and quantification. The aim is to unify all user experience-related data into calculable quantitative values, laying the foundation for subsequent experience scoring calculations.

[0046] After obtaining all quantified values, to more accurately reflect the impact of different types of data on user experience levels, a weighted average is applied to all quantified values ​​within the same technical field based on a pre-defined set of weighting coefficients. For example, in a specific technical field, the weight of completing complex repair tasks might be higher than that of consulting ordinary documents, and the weight of obtaining external professional certifications might be higher than that of internal training courses. These weighting coefficients can be set and optimized based on domain experts' experience, historical data analysis, or machine learning models. Through this weighted processing, a comprehensive experience score for the user in that technical field can be obtained, which can more comprehensively and objectively measure the user's professional competence in that field.

[0047] In practical applications, the experience score is a continuous numerical value. For ease of management and application, it is necessary to classify the user's experience level in the corresponding technical field into multiple preset level categories based on the experience score. For example, experience levels can be divided into "beginner," "intermediate," "advanced," or "expert." Each level category corresponds to an experience score range. When a user's experience score falls into a certain range, they are classified into the corresponding experience level. The purpose is to transform complex quantitative scores into intuitive and easy-to-understand level identifiers, facilitating subsequent risk assessment and operational suggestion generation by the system.

[0048] Furthermore, to ensure the real-time nature and accuracy of the domain-specific experience level profiles, the system repeatedly performs the aforementioned quantitative conversion, weighted summation, and level classification processes when it acquires new user action data, task execution data, training certification data, and evaluation feedback data. This means that a user's experience score and corresponding experience level category will be updated in real time as their latest actions and qualifications are updated. For example, when a user completes a new, challenging task or obtains a new professional certification, their experience score in the corresponding technical field will increase accordingly, and they may be promoted to a higher experience level.

[0049] Through the aforementioned technical solutions, this application can construct more accurate, comprehensive, and dynamically updated user domain-specific experience level profiles. These profiles not only effectively integrate multi-source heterogeneous data, including non-numerical information, but also ensure the objectivity and reliability of experience scoring through scientific quantification and weighting mechanisms. More importantly, the real-time updating mechanism ensures that the experience profiles always reflect the user's latest skill level, avoiding risk assessment biases caused by information lag. This provides a solid and reliable foundation for subsequent knowledge retrieval, risk level determination, and operational suggestion generation, significantly improving the accuracy and practicality of the entire knowledge base intelligent retrieval and management method.

[0050] In some preferred embodiments, a specific example is given below. Suppose a user named "Zhang San" is in the field of "equipment maintenance" technology, and their multi-dimensional behavioral and qualification data includes: Operational behavior data: Consulted the "Pump Maintenance Manual" 3 times (each time lasting an average of 10 minutes), and watched the "Valve Troubleshooting Video" once (lasting 30 minutes).

[0051] Task execution data: Completed "Emergency Repair Task for Type A Pump" 1 time (high difficulty), and completed "Regular Maintenance Task for Type B Valve" 2 times (medium difficulty).

[0052] Training and certification data: Completed internal "Safety Operating Procedures" training (completed), obtained external "Senior Mechanical Repair Technician" certification.

[0053] Evaluation feedback data: Expert evaluation: "Standardized operation, experienced" (positive evaluation); Colleague feedback: "Prompt problem solving" (positive feedback).

[0054] Specifically, the system first performs quantization transformation on this data: The time allotted for consulting the "Pump Maintenance Manual" (30 minutes) is quantified as 30 minutes.

[0055] The viewing time of the "Valve Troubleshooting Video" (30 minutes) is quantified as 30 minutes.

[0056] One successful “Emergency Repair Task for Type A Pump” (high difficulty) is worth 100 points.

[0057] The "Type B Valve Regular Maintenance Task" is completed twice (medium difficulty), which is quantified as 2 * 50 = 100 points.

[0058] The internal "Safety Operating Procedures" training is marked as "completed" and is quantified as 20 points.

[0059] The external "Senior Mechanical Repair Technician" certification is quantified as 150 points.

[0060] Experts rated the operation as "standardized and experienced" as 80 points.

[0061] Colleague collaboration feedback: "Quick problem-solving" was quantified as 70 points.

[0062] Next, a weighted average is calculated based on a pre-defined set of weighting coefficients. Assuming that in the "equipment maintenance" domain, task execution data has the highest weight (e.g., 0.4), followed by training and certification data (0.3), then operational behavior data (0.15) and evaluation feedback data (0.15). Zhang San's experience score in this domain would then be calculated as follows: Experience score = (30+30)*0.15 + (100+100)*0.4 + (20+150)*0.3 + (80+70)*0.15 = 60*0.15 + 200*0.4 + 170*0.3 + 150*0.15 = 9 + 80 + 51 + 22.5 = 162.5 points.

[0063] Subsequently, according to the preset experience level classification rules, for example: 0-100 points are "beginner", 101-200 points are "intermediate", and 201-300 points are "advanced". Zhang San's experience score of 162.5 points will classify his experience level in the "equipment maintenance" field as "intermediate".

[0064] When Zhang San subsequently completes another "Type C Equipment Fault Diagnosis" task (high difficulty, quantified as 120 points), the system will acquire this new data in real time and repeat the above-mentioned quantification conversion, weighted summation, and level classification processing. The new experience score will be recalculated, for example, becoming: New experience rating = (30+30)*0.15 + (100+100+120)*0.4 + (20+150)*0.3 + (80+70)*0.15 = 9 + 320*0.4 + 51 + 22.5 = 9 + 128 + 51 + 22.5 = 210.5 points.

[0065] At this point, Zhang San's experience level in the "Equipment Maintenance" field will be updated to "Advanced". In this way, the user's experience level profile can always be kept up-to-date and accurate.

[0066] In some embodiments described above in this application, a method is proposed to obtain searchable knowledge matching the search request from a searchable knowledge set and to determine the risk level based on the contextual attribute information of the searchable knowledge. Specifically, the generation method of the searchable knowledge set and the acquisition and definition of contextual attribute information can be further described as follows: The knowledge set to be retrieved is generated from multi-source heterogeneous data through unified indexing and attribute annotation. The metadata of each piece of knowledge contains pre-annotated contextual attribute information. The step of retrieving the knowledge set that matches the retrieval request includes: According to the retrieval request, knowledge with matching content is selected from the set of knowledge to be retrieved as the knowledge to be retrieved, and the context attribute information in the metadata of the knowledge to be retrieved is extracted. The scenario attribute information includes a first scenario attribute and a second scenario attribute. The first scenario attribute is used to characterize the knowledge as record-type content generated in a real business scenario, and the second scenario attribute is used to characterize the knowledge as simulated demonstration-type content generated in a teaching and training scenario. The scenario attribute information is used in conjunction with the sub-domain experience level file to determine the risk level.

[0067] Specifically, to ensure the effectiveness of the knowledge base and the accuracy of retrieval, the construction process of the knowledge set to be retrieved involves unified indexing and attribute annotation of multi-source heterogeneous data. Unified indexing refers to standardizing data from different sources and formats (such as document data, structured business data, video data, etc.) and establishing a unified retrieval structure, enabling all knowledge content to be retrieved efficiently. Attribute annotation refers to pre-adding contextual attribute information to the metadata of each piece of knowledge when it is added to the knowledge base. This attribute information plays a crucial role in subsequent risk assessment.

[0068] The process of retrieving knowledge that matches the search request from the knowledge set begins by using a unified indexing mechanism to filter out knowledge entries that highly match the search keywords or semantic content from the vast knowledge set based on the user's submitted search request. These filtered knowledge entries are the searchable knowledge. Subsequently, the system extracts pre-labeled contextual attribute information from the metadata of this searchable knowledge.

[0069] Furthermore, the contextual attribute information is subdivided into a first contextual attribute and a second contextual attribute. The first contextual attribute is used to identify record-type content generated in real business scenarios, such as actual operation logs, fault handling reports, and project implementation documents. This knowledge typically has high practicality and authenticity. The second contextual attribute is used to identify simulation demonstration-type content generated in teaching and training scenarios, such as training materials, simulation exercise videos, and theoretical explanation documents. This knowledge typically focuses on theoretical instruction and skills demonstration. By distinguishing between these two types of contextual attributes, the system can more precisely understand the source and nature of knowledge, thereby more accurately determining the risk level of the current user corresponding to the retrieved knowledge when subsequently combined with the user's domain-specific experience level profile.

[0070] Through the above technical solution, this application can construct a more structured and comprehensive set of knowledge to be retrieved, richer in information. By pre-indexing and annotating the knowledge with contextual attributes, especially distinguishing between record-type content generated in real business scenarios and simulation demonstration-type content generated in teaching and training scenarios, the system can not only efficiently obtain content-matching knowledge when responding to retrieval requests, but also accurately obtain its contextual information. Therefore, when subsequently determining the risk level of the user corresponding to the retrieved knowledge, a more comprehensive and refined judgment can be made based on knowledge attributes, significantly improving the accuracy and refinement of risk assessment and providing users with a safer and more suitable knowledge retrieval experience.

[0071] In some of the embodiments described above in this application, the risk level of the current user corresponding to the knowledge to be retrieved is determined based on the domain experience level profile and the contextual attribute information of the knowledge to be retrieved. However, if there is a lack of clear and unified judgment criteria in its implementation, the risk level determination process may be subjective or inconsistent, thereby affecting the accuracy and effectiveness of subsequent knowledge presentation and operational suggestions.

[0072] In this regard, this application further proposes the following steps for determining the risk level of the current user corresponding to the knowledge to be retrieved based on the domain-specific experience level profile and the contextual attribute information of the knowledge to be retrieved: Obtain a preset set of risk assessment rules, which includes the correspondence between multiple domain experience levels and scenario attributes.

[0073] Extract the experience levels from the domain-specific experience level archives, and match the experience levels with the contextual attribute information of the knowledge to be retrieved, and with the correspondence in the risk judgment rule set; Based on the matching results, the risk level of the current user corresponding to the knowledge to be retrieved is determined, and the risk level includes three levels: high risk, medium risk and low risk.

[0074] Specifically, the risk assessment rule set refers to a predefined series of logical rules, the core of which lies in establishing a correlation between domain-specific experience levels and contextual attributes to guide risk level assessment. For example, these rules can be defined as "when a user's experience level in a certain technical field is beginner, and the contextual attribute of the knowledge to be retrieved is record-type content generated in a real business scenario, then the knowledge is high-risk for that user." The correspondence can be stored and represented in the form of lookup tables, decision trees, or logical expressions to ensure the standardization and automation of the assessment process.

[0075] Extracting experience levels from the domain-specific experience level profile refers to obtaining the user's specific experience level within the technical field of the knowledge to be retrieved from the current user's domain-specific experience level profile, such as "beginner," "intermediate," or "advanced." The contextual attribute information of the knowledge to be retrieved may include a first contextual attribute (e.g., record-type content in real business scenarios) and a second contextual attribute (e.g., simulated demonstration-type content in teaching and training scenarios). These attributes together describe the source, nature, and scope of application of the knowledge.

[0076] Matching the experience level and the scenario attribute information with the correspondence in the risk assessment rule set means that the system, based on a preset matching algorithm, takes the current user's experience level and the scenario attribute of the knowledge to be retrieved as input, and searches for rules that meet the conditions in the risk assessment rule set. For example, if the rule set contains a rule that states "when the experience level is beginner and the scenario attribute is the first scenario attribute, the risk level is high risk," and the current input meets this condition, then the match is successful.

[0077] Therefore, based on the matching results, the risk level of the retrieved knowledge corresponding to the current user can be determined. The risk level is divided into three levels: high risk, medium risk, and low risk, aiming to quantitatively assess the potential impact of the knowledge content on the current user. High risk typically indicates a high risk of operational errors or misunderstandings for the current user; medium risk indicates some potential risk, requiring the user to treat it with caution; low risk indicates that the knowledge content poses a relatively low risk to the current user and can be used with relative peace of mind.

[0078] Through the above technical solution, this application enables objective, accurate, and automated assessment of the knowledge risk level in multi-source heterogeneous data knowledge bases. Compared to relying solely on user experience or knowledge attributes for judgment, this application combines both and matches them using preset rules, significantly improving the granularity and reliability of risk assessment. This effectively avoids potential operational risks or misunderstandings caused by a mismatch between user capabilities and knowledge content, ensuring that users receive risk warnings and guidance commensurate with their abilities when acquiring knowledge, thereby improving the usability and security of the knowledge base and optimizing the user experience in acquiring and applying knowledge.

[0079] In some preferred embodiments, a specific example is given below. Assume a company has a multi-source heterogeneous data knowledge base containing a large amount of knowledge about equipment maintenance. User Zhang San is a newly hired maintenance engineer whose experience level in the "hydraulic system maintenance" technical field is listed as "beginner". When Zhang San searches for "hydraulic pump troubleshooting", the system retrieves a searchable knowledge entry titled "Emergency On-site Repair Record of a Certain Model Hydraulic Pump" from the search knowledge set. The technical field of this knowledge is identified as "hydraulic system maintenance," and its scenario attribute information is "first scenario attribute," meaning it is a record-type content generated in a real business scenario.

[0080] At this point, the system will obtain a preset set of risk assessment rules. This set of rules includes one rule: "When the user's experience level in the field of 'hydraulic system repair' is 'beginner' and the scenario attribute of the knowledge to be retrieved is 'first scenario attribute', then the risk level of the knowledge is 'high risk'."

[0081] The system extracts Zhang San's experience level of "beginner" in the field of "hydraulic system repair" and the scenario attribute information "first scenario attribute" of the knowledge to be retrieved, "on-site emergency repair record of a certain type of hydraulic pump". The system then matches this information with the corresponding relationships in the risk assessment rule set. Since Zhang San's experience level and the scenario attribute of the knowledge perfectly match the above rules, the system ultimately determines that the risk level for Zhang San corresponding to the knowledge to be retrieved, "on-site emergency repair record of a certain type of hydraulic pump", is "high risk".

[0082] Based on this high-risk level, the system will dynamically adjust how the knowledge is presented in the search results. For example, it may add eye-catching visual warning labels and pop up a risk warning confirmation interface when Zhang San tries to access the knowledge. At the same time, it may recommend alternative knowledge that is more suitable for junior engineers, such as "Basic Principles Tutorial of Hydraulic Systems" or "Daily Maintenance Standard Procedures for Hydraulic Pumps".

[0083] In some of the above implementations, although the risk level of the knowledge to be retrieved can be determined based on the user's domain-specific experience level profile and the contextual attribute information of the knowledge to be retrieved, and the presentation of the search results and the generation of operation suggestions can be dynamically adjusted according to the risk level, if only the general presentation method is adjusted for knowledge content that is judged to be high-risk, it may not be sufficient to fully warn the user of the potential risks, especially when the user's experience level is low, there is a risk of misuse or misunderstanding of high-risk knowledge, which may lead to errors or adverse consequences in actual operation.

[0084] In response, this application further proposes operational suggestions for dynamically adjusting the presentation of the knowledge to be retrieved in the search results based on the risk level, and generating operational recommendations that match the domain-specific experience level profiles, including: When the risk level is high, a prominent visual warning label is added to the search results for the knowledge to be retrieved.

[0085] When a user triggers an access operation for the knowledge to be retrieved, a confirmation interface containing risk warning content pops up, and the confirmation interface includes a user confirmation control; Based on the topic identifier of the knowledge to be retrieved, a list of alternative knowledge recommendations related to the topic of the knowledge to be retrieved is generated and displayed. The list of alternative knowledge recommendations includes verified real business records, enterprise-certified standard procedures, and contact information of experts in the corresponding fields.

[0086] Specifically, the prominent visual warning sign can be understood as a visual element presented in a prominent manner in the search results. For example, it could be a red border, a flashing icon, a highlighted background color, or a warning label with the word "High Risk." Its purpose is to attract the user's attention immediately and clearly inform the user that the knowledge content has a high risk level. The risk warning confirmation interface refers to an interactive window that automatically pops up when a user attempts to access high-risk knowledge. This interface details the potential risks, applicable conditions, and possible consequences of the knowledge, aiming to force the user to fully understand and confirm the risks before accessing it. The user confirmation control can be a "I understand the risks and confirm access" button or checkbox; the user must actively perform this action to continue accessing the information, thereby ensuring the user's awareness and acceptance of the risks. In practical applications, the alternative knowledge recommendation list is a set of recommended content intelligently generated and displayed by the system based on the topic identifier of the knowledge to be searched. Its purpose is to provide users with safer, more reliable, or more suitable alternative information for their current experience level. The alternative knowledge recommendation list can include various types of knowledge. For example, verified real business records refer to cases, processes, or data that have been successfully applied and verified in actual business operations, possessing high reliability and practical guidance significance; enterprise-certified standard procedures refer to operation manuals, technical specifications, or safety procedures issued and formally certified by the enterprise's authoritative internal departments, representing the enterprise's best practices and mandatory requirements; and contact information for relevant field experts provides contact information for experts in related technical fields so that users can obtain professional support in a timely manner when encountering questions or needing further guidance.

[0087] Through the aforementioned technical solution, this application significantly improves the security and accuracy for users when processing high-risk knowledge. Prominent visual warning labels and a mandatory risk confirmation interface effectively prevent users from blindly accessing and using high-risk knowledge without understanding the risks, thereby reducing potential business risks and operational errors. Simultaneously, providing verified alternative knowledge and expert contact information not only offers users a safer and more reliable way to acquire knowledge but also guides them to gradually improve their knowledge comprehension and risk identification capabilities in specific technical fields, ultimately ensuring the effective utilization of the enterprise's internal knowledge base and the security of business operations.

[0088] In some preferred embodiments, suppose a newly hired equipment maintenance engineer's experience level profile in a specific area of ​​complex equipment repair is listed as "Junior". When the engineer searches for knowledge related to "high-voltage circuit repair", the system identifies that this knowledge content has a high operational risk, and its contextual attribute information may be "record-type content generated in a real business scenario". Combined with the engineer's "Junior" experience level, the system determines that the risk level of the knowledge to be searched for for the current user is "High Risk". At this time, in the search results, the "high-voltage circuit repair" knowledge entry will be marked with a red "High Risk Warning" tag, possibly accompanied by a flashing icon. When the engineer clicks on the entry, the system will pop up a confirmation interface, displaying "This operation involves high-voltage circuits and poses a risk of electric shock. Please be sure to perform this under the guidance of a professional. Do you confirm access?" and providing a "I understand and confirm access" button. Meanwhile, in the sidebar of the confirmation interface or search results page, the system will recommend an alternative knowledge list, which may include "High-voltage circuit safety operation procedures (enterprise certified)," "High-voltage equipment maintenance training videos (verified)," and "Contact information for electrical engineer Zhang (expert)." In this way, the system not only warns of risks but also provides engineers with a safer and more professional learning and operation path, effectively avoiding potential dangers caused by insufficient experience.

[0089] In some embodiments described above, this application proposes determining the risk level of the knowledge to be retrieved based on the user's domain-specific experience level profile and the contextual attribute information of the knowledge to be retrieved, and dynamically adjusting the presentation of the knowledge and generating operational suggestions based on the risk level. However, in practical applications, the processing strategies for knowledge with different risk levels need to be finely differentiated. In particular, when knowledge is judged to be of medium risk, if only strong restrictive measures similar to those for high-risk knowledge are adopted, it may excessively restrict the user's exploration and learning of the knowledge, which is not conducive to the improvement of the user's capabilities; on the other hand, if no risk warnings are given at all, users may misuse the knowledge without fully understanding it, thereby triggering potential business risks.

[0090] In response, this application further proposes that when the risk level is medium risk, an auxiliary prompt icon be added to the search results for the knowledge to be retrieved. The auxiliary prompt icon is used to prompt the user to use the knowledge to be retrieved in conjunction with the actual business scenario. A supplementary knowledge recommendation list related to the topic of the knowledge to be retrieved is generated and displayed. The supplementary knowledge recommendation list includes business manuals and similar business case information related to the content of the knowledge to be retrieved.

[0091] Specifically, the auxiliary prompts can be understood as visual or textual prompts designed to attract the user's attention in a non-mandatory manner within the search results. For example, the prompt could be a specific icon, a highlighted color, or a short text message, aiming to remind users to carefully judge and verify information in light of their actual business scenarios when consulting and applying the retrieved knowledge, avoiding blindly adopting information that may not be entirely applicable. This prompt aims to guide users to think critically, rather than directly preventing them from accessing the information. The supplementary knowledge recommendation list refers to the system intelligently recommending a series of related auxiliary knowledge resources to the user based on the topic and content of the retrieved knowledge. Unlike alternative knowledge that might be recommended in high-risk scenarios, the supplementary knowledge here aims to enrich the user's understanding of the depth and breadth of current medium-risk knowledge. Specifically, this list may include business manuals related to the retrieved knowledge content. These manuals typically provide standardized operating procedures, specifications, and background information, helping users fully understand the applicable boundaries and operational requirements of the knowledge. In addition, the list can also include information on similar business cases, providing users with specific application scenarios and lessons learned through actual successes or failures, helping users to better assess and avoid potential risks.

[0092] Through the aforementioned technical solution, this application provides a more refined and user-friendly approach to handling medium-risk knowledge to be retrieved. On one hand, the introduction of auxiliary prompts allows users to receive timely warnings when faced with knowledge that carries some risk but is not entirely unusable, prompting them to conduct necessary reflection and verification before application, thereby effectively reducing the risk of potential misuse. On the other hand, the provision of supplementary knowledge recommendation lists, particularly the integration of business manuals and similar business case information, greatly enriches the dimensions of knowledge acquisition for users, helping them build a more comprehensive knowledge system and improving their risk identification and operational execution capabilities in specific technical fields. This strategy avoids excessive restrictions on users' knowledge acquisition while effectively guiding them to utilize knowledge safely and efficiently, achieving a good balance between risk management and knowledge empowerment.

[0093] In some preferred embodiments, suppose a junior engineer working on equipment maintenance within an enterprise has an experience level of "Intermediate" in a specific equipment fault diagnosis area. When this engineer searches for "Handling method for fault code A101 of XX model equipment," the system filters out a relevant piece of knowledge from the search knowledge set. This knowledge's contextual attribute information shows it as "First Contextual Attribute," meaning it's a record-type content generated in a real business scenario, but its verification status is "Pending Verification." Based on preset risk assessment rules, the system classifies this searchable knowledge as "Medium Risk." At this point, in the search results list, a prominent yellow exclamation mark icon is added next to the "Handling method for fault code A101 of XX model equipment" knowledge entry as an auxiliary prompt, along with a brief message: "This knowledge content is recommended for careful reference in conjunction with actual business scenarios." Simultaneously, below the search results or in the sidebar, the system generates and displays a supplementary knowledge recommendation list. This list may contain the following: 1. "Standard Operating Manual for Maintenance of XX Model Equipment": Provides the official standard procedures and safety specifications for fault diagnosis and repair of this equipment.

[0094] 2. "Case Studies on Troubleshooting A101 Similar Equipment": This book includes successful experiences and precautions from other engineers when handling similar faults.

[0095] 3. "Advanced Troubleshooting Training Course for XX Equipment": Recommend relevant in-house training courses to enhance engineers' professional skills in this field.

[0096] In this way, when the engineer consulted the "Troubleshooting Method for Fault Code A101 of XX Model Equipment", he not only obtained a preliminary solution, but was also alerted by the system to potential risks and gained resources for further learning and verification, thus enabling him to conduct fault diagnosis and handling more comprehensively and safely.

[0097] In some embodiments described above, this application proposes determining the risk level of the knowledge to be retrieved based on the user's domain-specific experience level profile and the contextual attribute information of the knowledge to be retrieved, and dynamically adjusting the presentation of the knowledge in the search results according to the risk level. However, in practical applications, even if knowledge is determined to be low-risk, users may still need to understand its background information, verification status, or best practices when consulting and applying this knowledge to ensure correct understanding and effective application. Without providing this additional information, users may not be able to fully utilize the value of low-risk knowledge, or may still misunderstand it in specific contexts. Therefore, this application further proposes optimizing the presentation of the knowledge to be retrieved when the risk level is low-risk and providing additional information to improve the user's understanding and application efficiency of the knowledge.

[0098] When the risk level is low, the knowledge to be retrieved is displayed in a preset default presentation mode; in the display interface of the knowledge to be retrieved, background information and best practice links related to the knowledge to be retrieved are added, and the background information is used to explain the verification status and scope of application of the knowledge to be retrieved.

[0099] Specifically, when the system determines that the knowledge to be retrieved is low-risk for the current user based on a preset set of risk assessment rules, the knowledge will be displayed in the search results in a preset default presentation mode. The default presentation mode typically refers to a conventional display mode without prominent visual warnings or auxiliary prompts, such as displaying knowledge titles and summaries in standard fonts, colors, and layouts. Furthermore, to provide users with more comprehensive information support even in low-risk scenarios, this application adds background information and best practice links related to the knowledge to be retrieved to the display interface. The background information can be understood as supplementary explanations of the knowledge, specifically including its source, creation time, last update time, verification status (e.g., "verified," "pending verification," "draft"), and the specific business scenarios or scope to which it applies. Its purpose is to help users quickly understand the reliability and applicability of the knowledge. Best practice links refer to external or internal links pointing to verified operational processes, methods, or cases related to the knowledge topic that have proven to be efficient or optimized in actual business operations. Its purpose is to guide users to learn and refer to best practices within the industry or enterprise, thereby improving their operational efficiency and quality.

[0100] Through the aforementioned technical solutions, even when the knowledge to be retrieved is deemed low-risk, users can still obtain more comprehensive and instructive information. Specifically, by providing background information, users can quickly assess the reliability and applicability of the knowledge, thereby enhancing their trust in the knowledge and reducing the potential risk of misuse. Simultaneously, the introduction of best practice links provides users with a way to learn from and draw upon advanced experiences, helping them adopt more efficient and standardized methods in practice, thereby improving their work efficiency and business quality. This refined information presentation and suggestion mechanism effectively enhances the utilization value of low-risk knowledge and the user experience without increasing the user's cognitive burden.

[0101] In some preferred embodiments, suppose an employee, Xiao Li, searches for the knowledge "Equipment Daily Maintenance Checklist" in the knowledge base. Based on Xiao Li's experience level profile in the equipment maintenance field (e.g., Xiao Li's experience level in this field is "Advanced") and the contextual attribute information of the knowledge (e.g., the knowledge is "Standard Procedure Content," and the contextual attribute is the first contextual attribute), the system determines that the risk level of this knowledge for Xiao Li is low. At this time, the system will display the title and summary of the "Equipment Daily Maintenance Checklist" in the search results according to the default presentation method. When Xiao Li clicks to view the detailed content of this knowledge, in addition to displaying the checklist itself, the system will also add additional background information, such as "This checklist has been verified by the equipment expert team in January 2023 and is applicable to daily inspections of all models of A-series equipment," and provide a best practice link, such as "View the A-series equipment preventive maintenance best practice manual." Through this additional information, Xiao Li can not only obtain the required checklist content, but also understand the authority and scope of application of the checklist, and further learn more efficient maintenance methods, thereby ensuring the standardization and effectiveness of his maintenance operations.

[0102] In some of the embodiments described above in this application, a scheme for unified indexing of multi-source heterogeneous data to support knowledge retrieval was proposed. However, in practical applications, multi-source heterogeneous data may contain various complex types, such as video data. Without specialized content analysis and indexing of video data, its internal audio information and video content may not be effectively identified and utilized, thus limiting the retrieval accuracy and recall of video knowledge and failing to fully realize the value of video as a knowledge carrier. Therefore, this application further proposes a scheme for in-depth content analysis and index construction of video data within multi-source heterogeneous data to improve the knowledge base's retrieval capabilities for video content.

[0103] In response, this application further proposes that the aforementioned multi-source heterogeneous data includes document-type data, structured business data, and video-type data; the aforementioned method also includes: Perform preset content analysis and processing on the above video data to extract the audio and text information and key object information in the picture; Based on the aforementioned voice and text information and key object information in the video, a corresponding unified index is established for the aforementioned video data. When responding to a retrieval request, the unified index is used to filter out content-matching knowledge from multi-source heterogeneous data based on retrieval keywords and key objects, and generate the aforementioned knowledge to be retrieved.

[0104] Specifically, multi-source heterogeneous data can be understood as a collection of data from different sources and in different formats within an enterprise, encompassing various information carriers. Among them, document data can include various reports, manuals, rules and regulations, and other text files; structured business data can refer to tabular data, database records, etc., exported from business systems; while video data specifically refers to media files containing dynamic images and sounds, such as training videos, operation demonstration videos, and meeting recordings.

[0105] Furthermore, performing pre-defined content analysis on video data refers to using artificial intelligence technologies, such as Automatic Speech Recognition (ASR) and Computer Vision (CV), to automatically parse video content. Voice-to-text information refers to converting the audio content in the video into searchable text using speech recognition technology. Key visual object information refers to identifying important visual elements such as entities, scenes, and actions appearing in the video frame using computer vision technology. For example, in a video of equipment repair, audio-to-text phrases such as "engine fault diagnosis" and "replacing parts" can be extracted, and key visual objects such as "wrench," "screwdriver," and "specific model of equipment" can be identified.

[0106] Based on this, a unified index is established for the aforementioned video data, using the aforementioned voice and text information and key object information from the video. This unified index involves structurally storing the text and visual features extracted from the video and associating them with the video itself. Its purpose is to enable the system to efficiently filter multi-source heterogeneous data based on the user's input search keywords and / or key objects when responding to a user's search request, thereby accurately identifying and matching video knowledge containing relevant content and presenting it to the user as the searchable knowledge.

[0107] Through the aforementioned technical solution, this application significantly enhances the intelligent retrieval and management capabilities of a multi-source heterogeneous data knowledge base for video-related knowledge. Specifically, by deeply extracting and indexing the audio, text, and key objects in the video content, the internal information of the video knowledge is fully utilized, effectively solving the problem of the difficulty in accurately retrieving video content in traditional methods. This not only improves the accuracy and recall of retrieval results, enabling users to obtain the video knowledge they need more conveniently and comprehensively, but also enriches the retrieval dimensions of the knowledge base, allowing it to better support knowledge discovery and utilization based on multimodal information, thereby improving the overall intelligence level and user experience of the knowledge base system.

[0108] In some embodiments, reference is made to Figure 3This application proposes an intelligent retrieval and management system for a multi-source heterogeneous data knowledge base, comprising: a collection unit 301, used to continuously collect multi-dimensional behavioral and qualification data of users within an enterprise, and to construct and maintain user-specific experience level profiles based on the multi-dimensional behavioral and qualification data. These profiles characterize users' knowledge comprehension, risk identification, and operational execution capabilities in different technical fields, with each technical field corresponding to an independent experience level profile; and a matching unit 302, used to respond to user retrieval requests by retrieving knowledge matching the retrieval request from a set of knowledge to be retrieved. The set of knowledge to be retrieved includes multi-source heterogeneous data... All knowledge content stored after unified indexing includes a corresponding technical field identifier for each piece of knowledge to be retrieved. Extraction unit 303 is used to extract the current user's sub-domain experience level profile in the corresponding technical field based on the technical field identifier of the knowledge to be retrieved, and to determine the risk level of the knowledge to be retrieved for the current user based on the sub-domain experience level profile and the contextual attribute information of the knowledge to be retrieved. Generation unit 304 is used to dynamically adjust the presentation of the knowledge to be retrieved in the search results based on the risk level, and to generate operation suggestions that match the sub-domain experience level profile. The operation suggestions include supplementary knowledge recommendations related to the knowledge to be retrieved and adapted to the current user's capabilities.

[0109] The system disclosed in this application aims to address the problem that traditional knowledge base retrieval systems, when processing multi-source heterogeneous data, cannot effectively distinguish the contextual attributes of knowledge and fail to fully consider the differences in users' experience across different technical fields, resulting in a mismatch between retrieval results and users' capabilities and the contextual attributes of knowledge. For example, in knowledge base systems integrating multi-source heterogeneous data, including videos, existing retrieval technologies mainly rely on direct analysis of data content, ignoring the "contextual attributes" of the knowledge itself, i.e., the background and intent behind the information. The system cannot effectively distinguish between a live video recording a real fault and a simulated demonstration video used for teaching, even if the two highly overlap in content keywords and visual elements. This leads to confusion between real problem diagnosis information and simulated, idealized teaching information in the retrieval results, not only increasing the cognitive burden on users in filtering information, but more seriously, when users (especially inexperienced technicians) misapply solutions from teaching scenarios to real, complex production scenarios, they may make incorrect decisions due to ignoring scenario differences, leading to safety risks or economic losses.

[0110] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent retrieval and management of multi-source heterogeneous data knowledge bases, characterized in that, include: Continuously collect multi-dimensional behavioral and qualification data of users within the enterprise, and build and maintain user experience level profiles in different fields based on the multi-dimensional behavioral and qualification data. The experience level profiles in different fields are used to characterize the user's knowledge understanding ability, risk identification ability and operational execution ability in different technical fields. Each technical field corresponds to an independent experience level profile. In response to a user's search request, the system retrieves the search knowledge that matches the search request from the search knowledge set. The search knowledge set contains all knowledge content stored after multi-source heterogeneous data is indexed in a unified manner, and each piece of search knowledge contains a corresponding technical field identifier. Based on the technical field identifier of the knowledge to be retrieved, the current user's sub-field experience level profile in the corresponding technical field is extracted. Based on the sub-field experience level profile and the contextual attribute information of the knowledge to be retrieved, the risk level of the current user corresponding to the knowledge to be retrieved is determined. Based on the risk level, the presentation of the knowledge to be retrieved in the search results is dynamically adjusted, and operational suggestions matching the sub-domain experience level profile are generated. The operational suggestions include supplementary knowledge recommendations related to the knowledge to be retrieved and adapted to the current user's capabilities.

2. The intelligent retrieval and management method for multi-source heterogeneous data knowledge base according to claim 1, characterized in that, The continuous collection of multi-dimensional behavioral and qualification data of users within the enterprise includes: Collect user action behavior data in the enterprise's multi-source heterogeneous data knowledge base system. The action behavior data includes search records, document viewing time, and video viewing records. Each piece of action behavior data is associated with a corresponding technical field identifier. Collect user task execution data in the enterprise business system. The task execution data includes the type and quantity of maintenance tasks completed by the user. Each piece of the task execution data is associated with a corresponding technical field identifier. The system collects user training and certification data and evaluation feedback data. The training and certification data includes the completion status of internal training courses and external professional certification information. The evaluation feedback data includes structured evaluations from experts and collaborative feedback from peers. Each piece of training and certification data and evaluation feedback data is associated with a corresponding technical field identifier. The operational behavior data, the task execution data, the training and certification data, and the evaluation feedback data are collectively used as the multi-dimensional behavior and qualification data.

3. The intelligent retrieval and management method for multi-source heterogeneous data knowledge base according to claim 2, characterized in that, The process of constructing and maintaining user-specific experience level profiles based on the multi-dimensional behavioral and qualification data includes: Perform a preset quantization transformation on the non-numerical data in the multi-dimensional behavior and qualification data to obtain the corresponding quantized values; Based on a preset set of weighting coefficients, all quantitative values ​​in the same technical field are weighted to obtain the user's experience score in the corresponding technical field. Based on the experience rating, the user's experience level in the corresponding technical field is divided into multiple preset level categories; When new user action data, task execution data, training certification data, and evaluation feedback data are obtained, quantitative conversion processing, weighted summation processing, and level classification processing are repeatedly performed to update the user's experience score and corresponding experience level category in the corresponding technical field in real time.

4. The intelligent retrieval and management method for multi-source heterogeneous data knowledge base according to claim 1, characterized in that, The set of knowledge to be retrieved is generated from multi-source heterogeneous data through unified indexing and attribute labeling. The metadata of each piece of knowledge contains pre-labeled contextual attribute information. The step of obtaining the knowledge to be retrieved from the knowledge set that matches the retrieval request includes: According to the retrieval request, knowledge with matching content is selected from the set of knowledge to be retrieved as the knowledge to be retrieved, and the context attribute information in the metadata of the knowledge to be retrieved is extracted. The scenario attribute information includes a first scenario attribute and a second scenario attribute. The first scenario attribute is used to characterize the knowledge as record-type content generated in a real business scenario, and the second scenario attribute is used to characterize the knowledge as simulated demonstration-type content generated in a teaching and training scenario. The scenario attribute information is used in conjunction with the sub-domain experience level file to determine the risk level.

5. The intelligent retrieval and management method for multi-source heterogeneous data knowledge base according to claim 1, characterized in that, The step of determining the risk level of the current user corresponding to the knowledge to be retrieved based on the domain-specific experience level profile and the contextual attribute information of the knowledge to be retrieved includes: Obtain a preset set of risk assessment rules, which includes a correspondence between multiple domain experience levels and scenario attributes; Extract the experience levels from the domain-specific experience level archives, and match the experience levels with the contextual attribute information of the knowledge to be retrieved, and with the correspondence in the risk judgment rule set; Based on the matching results, the risk level of the current user corresponding to the knowledge to be retrieved is determined, and the risk level includes three levels: high risk, medium risk and low risk.

6. The intelligent retrieval and management method for multi-source heterogeneous data knowledge base according to claim 5, characterized in that, The step of dynamically adjusting the presentation of the knowledge to be retrieved in the search results based on the risk level, and generating operational suggestions that match the sub-domain experience level profile, includes: When the risk level is high, a prominent visual warning label is added to the search results for the knowledge to be retrieved; When a user triggers an access operation for the knowledge to be retrieved, a confirmation interface containing risk warning content pops up, and the confirmation interface includes a user confirmation control; Based on the topic identifier of the knowledge to be retrieved, a list of alternative knowledge recommendations related to the topic of the knowledge to be retrieved is generated and displayed. The list of alternative knowledge recommendations includes verified real business records, enterprise-certified standard procedures, and contact information of experts in the corresponding fields.

7. The intelligent retrieval and management method for multi-source heterogeneous data knowledge base according to claim 5, characterized in that, The step of dynamically adjusting the presentation of the knowledge to be retrieved in the search results based on the risk level, and generating operational suggestions that match the sub-domain experience level profile, includes: When the risk level is medium risk, an auxiliary prompt icon is added to the search results for the knowledge to be searched. The auxiliary prompt icon is used to prompt the user to use the knowledge to be searched in combination with the actual business scenario. Generate and display a supplementary knowledge recommendation list related to the knowledge topic to be retrieved. The supplementary knowledge recommendation list includes business manuals and similar business case information related to the knowledge content to be retrieved.

8. The intelligent retrieval and management method for multi-source heterogeneous data knowledge base according to claim 5, characterized in that, The step of dynamically adjusting the presentation of the knowledge to be retrieved in the search results based on the risk level, and generating operational suggestions that match the sub-domain experience level profile, includes: When the risk level is low, the knowledge to be retrieved is displayed according to the preset default presentation method; In the display interface of the knowledge to be retrieved, add background information and best practice links related to the knowledge to be retrieved. The background information is used to explain the verification status and scope of application of the knowledge to be retrieved.

9. The intelligent retrieval and management method for multi-source heterogeneous data knowledge base according to claim 1, characterized in that, The multi-source heterogeneous data includes document data, structured business data, and video data; the method further includes: Perform preset content analysis processing on the video data to extract the audio and text information and key object information in the video; Based on the voice and text information and the key object information in the video, a corresponding unified index is established for the video data. When responding to a retrieval request, the unified index is used to filter out content-matching knowledge from multi-source heterogeneous data based on retrieval keywords and key objects, and generate the knowledge to be retrieved.

10. A multi-source heterogeneous data knowledge base intelligent retrieval and management system, characterized in that, include: The data collection unit is used to continuously collect multi-dimensional behavioral and qualification data of users within the enterprise, and to build and maintain user experience level profiles in different fields based on the multi-dimensional behavioral and qualification data. The experience level profiles in different fields are used to characterize the user's knowledge understanding ability, risk identification ability and operational execution ability in different technical fields. Each technical field corresponds to an independent experience level profile. The matching unit is used to respond to the user's search request and retrieve the search knowledge that matches the search request from the search knowledge set. The search knowledge set contains all knowledge content stored after multi-source heterogeneous data is uniformly indexed. Each piece of search knowledge contains a corresponding technical field identifier. The extraction unit is used to extract the current user's sub-domain experience level profile in the corresponding technical field based on the technical field identifier of the knowledge to be retrieved, and to determine the risk level of the current user corresponding to the knowledge to be retrieved based on the sub-domain experience level profile and the contextual attribute information of the knowledge to be retrieved. The generation unit is used to dynamically adjust the presentation of the knowledge to be retrieved in the search results according to the risk level, and generate operation suggestions that match the sub-domain experience level profile. The operation suggestions include supplementary knowledge recommendations related to the knowledge to be retrieved and adapted to the current user's capabilities.

Images