Conference data management system and method based on artificial intelligence

Abstract

The invention discloses a conference data management system and method based on artificial intelligence, and relates to the technical field of data management, and the method comprises the steps: determining a distributed conference database, and appointing a management demand of conference data; the method comprises the following steps: decomposing management requirements, determining in-library management requirements for one-step training, determining inter-library overall planning requirements for two-step training, and generating a library management unit and an upper computer unit; calling a library management unit, transferring the library management unit to a conference database to carry out multiple rounds of internal competition training based on local library characteristics, and determining a local library management unit; establishing connection between the local library management unit and the upper computer unit to form a large management model; and receiving a conference data management task, and performing task management in combination with the management large model. The technical problems of inaccurate task allocation, low cooperation efficiency and insufficient data processing capability in distributed conference data management in the prior art are solved, and the technical effect of improving the conference data management efficiency and the cooperation capability is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of data management, and in particular to a conference data management system and method based on artificial intelligence. Background Art

[0002] With the rapid development of information and intelligent technology, conference data management has been widely used in corporate office and academic exchanges. However, with the rapid growth of data volume and the diversification of management needs, the storage, retrieval and task allocation process of conference data are facing increasing challenges. Traditional conference data management methods often have problems such as inaccurate task allocation, low collaboration efficiency, and insufficient data processing capabilities, which are difficult to meet the needs of modern distributed conference systems for efficient, accurate and reliable management. Summary of the invention

[0003] The present application provides an artificial intelligence-based conference data management system and method, which is used to solve the technical problems of inaccurate task allocation, low collaboration efficiency and insufficient data processing capability in the prior art in distributed conference data management.

[0004] In view of the above problems, the present application provides a conference data management system and method based on artificial intelligence.

[0005] The first aspect of the present application provides a conference data management system based on artificial intelligence, the system comprising:

[0006] A management demand determination module, the management demand determination module is used to determine a distributed conference database and specify the management demand of conference data, wherein the conference database includes multiple ones and is marked with a unique library code; a first training module, the first training module is used to decompose the management demand, determine the management demand within the library for one-step training, determine the coordination demand between libraries for two-step training, and generate a library management unit and a host unit; a second training module, the second training module is used to call the library management unit, delegate it to the conference database for multiple rounds of internal competitive training based on local library characteristics, and determine the local library management unit, wherein the local library management unit corresponds to the conference database one by one and is built into the conference database; a management large model construction module, the management large model construction module establishes a connection between the local library management unit and the host unit based on the library interface to form a management large model, wherein the library interface is marked with a library code; a task management module, the task management module receives conference data management tasks, and performs task management in combination with the management large model, wherein task management includes subtask allocation based on the host unit and multi-library parallel processing based on the local library management unit.

[0007] The second aspect of the present application provides a conference data management method based on artificial intelligence, the method comprising:

[0008] Determine a distributed conference database and specify the management requirements of conference data, wherein the conference database includes multiple databases and is identified with a unique database code; decompose the management requirements, determine the management requirements within the database for one-step training, determine the coordination requirements between the databases for two-step training, and generate a database management unit and a host unit; call the database management unit and delegate it to the conference database for multiple rounds of internal competitive training based on local database characteristics to determine the local database management unit, wherein the local database management unit corresponds to the conference database one-to-one and is built into the conference database; establish a connection between the local database management unit and the host unit based on the library interface to form a large management model, wherein the library interface is marked with a library code; receive conference data management tasks, and perform task management in combination with the large management model, wherein task management includes subtask allocation based on the host unit and multi-library parallel processing based on the local database management unit.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] The present application determines a distributed conference database and specifies the management requirements of conference data, wherein the conference database includes multiple conference databases and is marked with a unique library code; the management requirements are decomposed to determine the management requirements within the library for one-step training, and the coordination requirements between libraries are determined for two-step training, and a library management unit and a host unit are generated; the library management unit is called and delegated to the conference database for multiple rounds of internal competitive training based on local library characteristics to determine a local library management unit, wherein the local library management unit corresponds to the conference database one-to-one and is built into the conference database; based on the library interface, a connection is established between the local library management unit and the host unit to form a management model, wherein the library interface is marked with a library code; conference data management tasks are received, and task management is performed in combination with the management model, wherein task management includes sub-task allocation based on the host unit and multi-library parallel processing based on the local library management unit. The present invention solves the technical problems of inaccurate task allocation, low coordination efficiency and insufficient data processing capability in the prior art in distributed conference data management. Through distributed database design, a large management model is constructed using a unique library coding identifier and a library interface. A library management unit and a host unit are generated through one-step training and two-step training. Multiple rounds of internal competitive training are performed in combination with local library characteristics to optimize the local library management unit, realize accurate task allocation and multi-library parallel processing, and achieve the technical effect of improving the efficiency and coordination capability of conference data management. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A schematic diagram of the structure of a conference data management system based on artificial intelligence provided in an embodiment of the present application;

[0013] Figure 2 A flowchart of a conference data management method based on artificial intelligence is provided in an embodiment of the present application.

[0014] Explanation of the reference numerals: management requirement determination module 11, first training module 12, second training module 13, management large model construction module 14, task management module 15. DETAILED DESCRIPTION

[0015] The present application provides an artificial intelligence-based conference data management system and method to solve the technical problems of inaccurate task allocation, low collaboration efficiency and insufficient data processing capability in the prior art in distributed conference data management. Through distributed database design, a large management model is constructed using unique library coding identifiers and library interfaces. Library management units and upper unit units are generated through one-step training and two-step training. Multiple rounds of internal competitive training are performed in combination with local library characteristics to optimize the local library management unit, realize accurate task allocation and multi-library parallel processing, and achieve the technical effect of improving the efficiency of conference data management and collaboration capability.

[0016] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0017] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0018] Embodiment 1, as Figure 1 As shown, the embodiment of the present application provides a conference data management system based on artificial intelligence, the system comprising:

[0019] The management requirement determination module 11 is used to determine a distributed conference database and specify the management requirements of conference data, wherein the conference database includes multiple ones, each of which is marked with a unique database code.

[0020] In the embodiment of the present application, the management demand determination module first automatically scans the conference database in the distributed environment through the network protocol or configuration file to identify all available database nodes. These database nodes together form a distributed architecture for collaborative storage and management of conference data. To ensure the uniqueness and identifiability of each database, a unique library code is assigned to each database, which is usually generated based on the network address, logical sequence or storage location of the database.

[0021] Next, the management requirements determination module specifies the management requirements for conference data in combination with the preset global management logic. These management requirements are unified management rules for all database nodes, defining the operational specifications for storage, classification, permission control, and data backup of conference data. Typical management requirements include document storage and classification, that is, classifying conference data according to common logic such as "first document" and "second document" to ensure the consistency of data structure; version control and update, that is, defining the version control rules for conference documents to ensure the synchronization of version updates in a multi-database collaborative environment; data access rights, that is, formulating common permission control logic to ensure that users follow a unified security policy when accessing distributed databases.

[0022] By associating common management requirements with distributed databases one by one and using library codes as identifiers, global management of distributed conference databases can be achieved.

[0023] The first training module 12 is used to decompose the management requirements, determine the management requirements within the warehouse for one-step training, determine the coordination requirements between warehouses for two-step training, and generate a warehouse management unit and a host unit.

[0024] In the embodiment of the present application, the first training module first analyzes and decomposes the received management requirements, and divides the requirements into intra-database management requirements and inter-database coordination requirements through a logical rule classification algorithm. Among them, intra-database management requirements focus on the localized operations of a single database, including data storage, classification, and permission control, and have universal and standardized characteristics; while inter-database coordination requirements are oriented to the collaborative management of multiple databases, including task allocation, data synchronization, and global optimization, emphasizing the collaboration of cross-database operations.

[0025] One-step training is performed for the determined in-library management requirements. By extracting the common rules of in-library requirements, standardized management logic is generated, and historical data management records are called for data integration to form standardized samples. Using these samples, a sample-driven training model is adopted, and through iterative training until convergence, a library management unit focusing on single-library operations is generated to ensure the efficiency and accuracy of single-database tasks.

[0026] Subsequently, two-step training is performed to meet the needs of inter-library coordination. First, multiple management goals are set, and these goals are initially coordinated and trained through reinforcement learning methods to generate the first version of the host. In order to improve the global coordination ability of the host, a dynamic adjustment strategy is further adopted to randomly update the management goals, and the host is iteratively optimized through transfer learning methods, and finally a fully functional host unit is generated.

[0027] Furthermore, in the system provided in the embodiment of the application, the first training module 12 is also used for:

[0028] Based on the management requirements within the library, determine the standardized management logic, wherein the standardized management logic is universal to the library; call the data management record, perform record integration based on the standardized management logic, and determine the standardized sample; based on the standardized sample, perform sample-driven training until convergence to generate the library management unit.

[0029] In the embodiment of the present application, the standardized management logic is first determined. The management requirements in the library are parsed by using a logic rule engine (such as Drools) to extract the core rules. By analyzing the characteristics of the requirements, including data storage format, classification rules and permission settings, a standardized management logic with universality is constructed. These logic rules are saved in the form of structured configuration files (such as JSON or XML).

[0030] The operation log and database snapshot of the historical database are then called to extract data management records. A database snapshot is a copy that captures the complete state of the database at a specific point in time. Combined with the operation log, it can truly reflect the management behavior of the database. The extracted data is preprocessed by using data cleaning technology, including removing outliers, removing redundant information, and repairing missing values ​​to ensure data consistency and integrity. Then, the cleaned data records are aligned and integrated based on standardized management logic to generate high-quality standardized samples.

[0031] Next, training is performed based on standardized samples, and the random forest algorithm is used to process the data. During the training process, the standardized samples are divided into a training set and a validation set. The training set is used to optimize the model so that it can accurately simulate the operational behavior of the database. The validation set is used to evaluate the performance of the model, using the mean square error (MSE) as the evaluation indicator. The smaller the mean square error, the smaller the deviation between the model prediction result and the actual value. When the mean square error remains stable and below the preset threshold in multiple consecutive iterations, the model is judged to have reached a convergence state.

[0032] After the model is trained and converged, it is combined with standardized management logic to generate an independent library management unit.

[0033] Furthermore, in the system provided in the embodiment of the application, the first training module 12 is also used for:

[0034] Set multiple management targets, perform coordinated training under the multiple management targets, and determine the first host device; perform random updates on the management targets, perform update management training under the multiple management targets on the first host device, and generate the host device unit.

[0035] In the embodiment of the present application, firstly, preset multiple management goals are set, including cross-database task allocation, data synchronization, load balancing, access control, etc. These goals are global management requirements for the collaboration needs between multiple databases, and serve as the core guiding rules for overall training.

[0036] Next, we perform coordinated training under multiple management objectives. We use reinforcement learning techniques (such as deep Q-learning) to optimize strategies by modeling the task and resource relationships between databases. First, we define the state space, taking the task queue length, storage capacity utilization, and load level of each database as state variables; and define the action space, including strategies for allocating tasks from one database to another, operations for synchronizing specific data, etc. To evaluate the effectiveness of the strategy, we establish a reward mechanism, and the reward value is calculated in combination with the degree of achievement of multiple management objectives. Specifically, the reward value is obtained by multiplying the task delay improvement, resource utilization improvement, and data consistency improvement by the corresponding preset weights. The task delay improvement is calculated as (initial delay time − current delay time) / initial delay time. The resource utilization improvement is quantified as the percentage reduction of the load standard deviation through the uniformity evaluation of load balancing. The data consistency improvement is calculated by the synchronization success rate × (1 − synchronization delay / maximum allowed delay), where / maximum allowed delay is preset. In the simulation environment, by continuously trying different task allocation and resource scheduling strategies, the strategy parameters are adjusted to gradually improve the efficiency of global task management. In this process, task allocation and scheduling logic are dynamically optimized based on the real-time calculated reward value to ensure performance balance under different management objectives. Finally, the first upper controller is generated through multiple rounds of strategy optimization.

[0037] To further enhance dynamic adaptability, the first host is trained for update management by introducing a random update mechanism to adapt to the dynamic changes in the number or characteristics of databases in actual scenarios. The random update mechanism generates random perturbations through Gaussian distribution, dynamically adjusts the number of databases, storage capacity, and task queue characteristics, and simulates the scenario of adding or removing nodes. The state space is expanded to include the storage capacity and task load of the newly added nodes, and the action space is updated to support the task initialization strategy of the newly added nodes and the load redistribution strategy after the nodes are removed. Update training re-optimizes the management logic of the host in a dynamic environment, and the reward mechanism introduces dynamic adaptability evaluation, such as the efficiency of task allocation of new nodes and the time to restore load balancing after removing nodes, to ensure that the host has the ability to quickly adapt to environmental changes. Through multiple rounds of optimization training, management logic that can handle dynamic environmental changes is generated.

[0038] Finally, a host unit with adaptive capabilities is generated. With the help of transfer learning technology, rapid adjustments are made based on the existing model parameters of the first host, and only the states and actions of newly added or removed nodes are trained, which significantly reduces the training time. After multiple rounds of optimization, the final host unit shows excellent performance in a dynamic multi-database environment.

[0039] The second training module 13 is used to call the library management unit, delegate it to the conference database to perform multiple rounds of internal competitive training based on local library characteristics, and determine the local library management unit, wherein the local library management unit corresponds one-to-one to the conference database and is built into the conference database.

[0040] In an embodiment of the present application, the second training module is used to call the library management unit and delegate it to each conference database, and generate a local library management unit corresponding to the conference database one by one through multiple rounds of internal competitive training based on local library characteristics. Specifically, the library management unit is delegated to the target conference database, and the local library characteristics of the database are called, such as storage type, access mode, and capacity characteristics. Based on these local characteristics, an initial learning rate is set for each characteristic, and a round of training is performed on the library management unit using the initial learning rate to generate a round of management units.

[0041] Subsequently, a round of management units is verified, and the learning rate is adjusted according to the local management standard, focusing on optimizing the underperforming parts. Multiple rounds of iterative training are carried out through the internal competition mechanism to continuously improve the adaptability of the management unit to the local characteristics. Finally, when the training reaches the preset standard, a local library management unit is generated. The local library management unit is closely integrated with the corresponding conference database and is built into the database to perform efficient storage, classification, and permission management operations.

[0042] Furthermore, in the system provided in the embodiment of the application, the second training module 13 is also used for:

[0043] The library management unit is decentralized to a first conference database, and a first local library feature of the first conference database is called, where the first conference database is any conference database; an initial learning rate is determined for the first local library feature, wherein each local library feature corresponds to an initial learning rate; based on the initial learning rate, a round of training is performed on the library management unit to determine a round of management units; for the round of management units, an internal adjustment of the learning rate based on the local management standard and multiple rounds of iterative training are performed to generate a first local library management unit.

[0044] In the embodiment of the present application, the library management unit is first decentralized to the first conference database, which can be any target database in the distributed system. As a general logic module, the library management unit is deployed to the database environment and calls the local features of the first conference database. Local library features include access mode (such as random access or sequential access), storage structure (such as file type or block type storage), capacity distribution and task load. These features are extracted by analyzing operation logs, parsing metadata tables and statistically analyzing historical operation records, and are integrated into feature vectors that describe database characteristics.

[0045] For the first local library feature extracted, set the initial learning rate. The learning rate is set using a simple linear normalization method. After normalizing the actual value of each feature to the range of 0 to 1, it is multiplied by a fixed scaling factor (such as 0.01) to generate a learning rate vector corresponding to each feature. This method ensures the transparency and feature relevance of the learning rate, and provides an appropriate optimization step for subsequent training.

[0046] Based on the calculated initial learning rate, the library management unit is trained for the first round, and the logic parameters of the library management unit are adjusted in combination with the first local library characteristics to make it initially adapt to the environment of the first conference database. After a round of training is completed, a round of management units is generated, and its logic is optimized to better adapt to the characteristics of the target database.

[0047] Then, for a round of management units, internal adjustment of the learning rate based on local management standards and multiple rounds of iterative training are performed to further improve the adaptability of the management unit. Internal adjustment of the learning rate refers to dynamically adjusting the learning rate corresponding to the feature according to the local management standard. For example, for features with insufficient performance, the learning rate is fine-tuned to 110% of the original value to enhance the optimization effect; for features with good performance, the learning rate remains unchanged. In each round of training, the logic of the management unit is further optimized in combination with the adjusted learning rate.

[0048] Through repeated optimization and internal competition adjustment, multiple rounds of iterative training continuously improve the adaptability of the management unit to the characteristics of the target database. When the training reaches the preset convergence condition, the training is considered complete and the first local database management unit is generated.

[0049] Furthermore, in the system provided in the embodiment of the application, the second training module 13 is also used for:

[0050] The first-round management unit is verified, and a management evaluation is performed based on the local management standard to determine an evaluation matrix, wherein the evaluation matrix has local library features as matrix rows, and feature management coefficients and fusion management coefficients as digital columns, and the fusion management coefficient is the degree of fusion between the standardized management logic and the first local library features; the evaluation matrix is ​​used to perform internal competition adjustment between learning rates, and two rounds of training are performed to determine the second-round management unit; for the second-round management unit, management evaluation and iterative training are performed until the preset convergence conditions are met to generate the first local library management unit.

[0051] In an embodiment of the present application, a round of management units is first verified, and their performance on the first conference database is evaluated according to preset local management standards. Local management standards include key performance indicators such as storage efficiency (such as the average time consumed for data writing), access latency (such as the time interval from request to response), and retrieval accuracy (such as the matching rate between search results and expected results). The data required for verification is obtained through performance sampling tools (such as database log analysis tools or query analysis tools). For example, storage efficiency data is obtained by counting the average time consumed for write operations, access latency is calculated by recording the time interval between the start time of each request and the completion time of the response, and retrieval accuracy is calculated by comparing the query results with the expected results.

[0052] The verification results are presented in the form of an evaluation matrix. The rows of the evaluation matrix represent the local features of the database (such as storage type, access mode, and capacity distribution), and the columns include feature management coefficients and fusion management coefficients. The feature management coefficient represents the degree of adaptation of the management unit to a specific local feature. It is calculated by comparing the deviation between the actual performance and the target performance. The feature management coefficient = 1-the absolute difference between the actual performance and the target performance divided by the target performance. For example, when the target access delay is 100 milliseconds and the actual delay is 120 milliseconds, the feature management coefficient is 1-(120-100) / 100=0.8, indicating that the degree of adaptation is 80%. The fusion management coefficient represents the degree of matching between the standardized management logic and the local features. It is calculated based on the performance of the standardized logic on the local features and the importance score of the local features. The fusion management coefficient = the performance of the standardized management logic divided by the importance of the local features. For example, if the performance of the standardized logic is 95% and the importance score of the local features is 0.9, the fusion management coefficient is 95 / 0.9=105.56, indicating that the standardized logic has a strong adaptability to the specific features.

[0053] Based on the evaluation matrix, internal competition between learning rates is adjusted to optimize the pertinence of the training process. Internal competition between learning rates is to dynamically optimize the learning rate corresponding to each feature based on the results of the evaluation matrix. For features with low feature management coefficients, slightly increase their learning rates, for example, increase the current learning rate by 10%, and the new learning rate = original learning rate × 1.1. For features whose fusion management coefficients have reached the standard, the learning rate remains unchanged. By adjusting the learning rate vector, ensure that training resources are first concentrated on features with insufficient performance, while avoiding over-optimization of parts that have reached the standard.

[0054] The adjusted learning rate is used for the second round of training. The updated learning rate vector is loaded, and the management logic is further optimized in combination with the parameters of the first round of management units. During training, the weighted average error function is used to optimize the parameters with the goal of minimizing the deviation of each performance indicator, and the objective function = the weighted average of the errors of all performance indicators. Among them, the weight of each performance is pre-set by the system according to the actual needs of database management. For example, for the three indicators of storage efficiency, access delay and retrieval accuracy, if it is believed that the access delay has the greatest impact on the overall performance, a higher weight (such as 0.5) can be assigned to it, while storage efficiency and retrieval accuracy are assigned lower weights (such as 0.3 and 0.2), respectively. By updating the parameters multiple times, a second round of management units are generated, whose logic is more adapted to the characteristics of the target database than the first round of management units.

[0055] For the second round of management units, continue to optimize the logic through management evaluation and iterative training. Each round of training repeats the process of verification, evaluation and learning rate adjustment to gradually improve the performance and adaptability of the management unit. When the feature management coefficient and fusion management coefficient in the evaluation array reach the preset standard, the training is considered complete and the final first local library management unit is generated.

[0056] The management large model construction module 14 establishes the connection between the local library management unit and the host unit based on the library interface to form a management large model, wherein the library interface is marked with library code.

[0057] In an embodiment of the present application, the management model construction module connects the local library management unit with the host unit through the library interface to form a management model to achieve unified management and efficient collaboration. The library interface adopts a standardized protocol design, corresponds to each local library management unit one by one, and is marked by a library code to ensure the uniqueness and accuracy of the interface. The library code is generated by the attributes of the database node and is used to identify the mapping relationship between the library interface and the local library management unit. The host unit accesses the status data and configuration information of each local library management unit through the library interface, and at the same time decomposes the global task into subtasks and distributes them to the target local library management unit for execution. The final management model combines the optimization capabilities of the local library management unit with the global scheduling capabilities of the host unit to provide efficient and unified management capabilities for distributed databases.

[0058] The task management module 15 receives the conference data management task and performs task management in combination with the management model, wherein the task management includes subtask allocation based on the host unit and multi-library parallel processing based on the local library management unit.

[0059] In an embodiment of the present application, after receiving the conference data management task, the task management module combines the management model to perform efficient task management. Specifically, the task management module first decomposes the received task through the upper unit, divides the task into multiple subtasks, and reasonably distributes them to the target database according to the real-time status of each local library management unit (such as load, capacity, response speed, etc.). Subsequently, the subtask is sent to the corresponding local library management unit through the library interface. The local library management unit independently executes the assigned subtasks in a multi-library parallel manner, and feeds back the task progress and results to the upper unit in real time.

[0060] Furthermore, in the system provided in the embodiment of the application, the task management module 15 is also used for:

[0061] The conference data management task is uploaded to the upper device unit, task allocation based on library storage type and library code matching are performed to determine the management subtask; based on the library interface, the management subtask is sent to the target local library management unit, and the library management of the subtask is executed in parallel.

[0062] In the embodiment of the present application, after uploading the conference data management task to the upper unit, the task content is first parsed, and the task is decomposed and allocated in combination with the information in the management model. Task uploading is implemented through the task receiving interface. After receiving the task, the upper unit parses its type (such as storage, retrieval, update, etc.) and specific requirements to prepare for the subsequent task decomposition.

[0063] Then, tasks are allocated based on the library storage type. The host unit obtains the storage type (such as file storage, block storage) and its adaptability information of each local library management unit by calling the management model, and decomposes the task into multiple management subtasks based on task requirements (such as the size and format of the data type in the storage task). Each subtask includes the data operation type and the adapted storage type.

[0064] Then, the target execution node is determined for each management subtask through library code matching. The library code is a unique identifier for each database node in a distributed system, which is used to accurately locate the execution node of the management subtask. The library interface mapping table is extracted from the management model, and the target database node of each management subtask is quickly matched through the library code to ensure the accuracy and efficiency of task allocation. After completing the task allocation and library code matching, the management subtask is sent to the target local library management unit through the library interface.

[0065] After the task is issued, each local library management unit independently executes the received subtasks and processes the assigned task content in a multi-library parallel manner. For example, different data blocks in the storage task can be distributed to different database nodes at the same time, and each node completes the storage operation separately. During this process, each local library management unit reports the task execution progress and results to the upper unit in real time through the library interface, which is convenient for global monitoring and dynamic adjustment.

[0066] Finally, by executing the library management subtasks in parallel, all management subtasks are completed efficiently across multiple nodes.

[0067] Furthermore, the system provided in the application embodiment is also used for:

[0068] If it is a storage task type, execute forward task processing based on the conference data management task and the newly added conference data, wherein the host unit-local library management unit is the forward direction; if it is a calling task type, execute forward task processing of the conference data management task and reverse calling processing of the stored conference data, wherein the host unit is called through the library interface to perform reverse allocation and interface display, and the library interface can perform two-way data interaction; if it is a transfer task type, execute forward task processing of the conference data management task, reverse calling processing based on the stored conference data, and forward transfer processing based on the stored conference data.

[0069] In an embodiment of the present application, if it is a storage task type, forward task processing based on the conference data management task and the newly added conference data is performed. Forward task processing refers to the operation process of transferring data from the upper device unit to the local library management unit. Specifically, after the upper device unit receives the storage task, it sends a storage request to the target local library management unit through the library interface, carrying the newly added conference data (such as meeting minutes, attachments, etc.). After receiving the data, the target local library management unit stores it in the specified location and records relevant metadata (such as storage time and path). After the storage is completed, the local library management unit feeds back the storage completion status to the upper device unit through the library interface. Through this process, the forward data transfer from the upper device unit to the local library management unit is realized to complete the storage task.

[0070] If it is a call task type, the forward task processing of the conference data management task and the reverse call processing of the stored conference data are executed. Forward task processing refers to passing the data call request from the host unit to the local library management unit, while the reverse call processing returns the retrieval result to the host unit through the library interface. Specifically, the host unit sends a call request to the target local library management unit through the library interface (such as querying the conference data of a specific time period). The local library management unit retrieves the stored data according to the request conditions and generates the retrieval results; the retrieval results are then passed back to the host unit through the library interface to complete the reverse allocation. The host unit formats the retrieval results and displays them on the user interface. This process realizes the reverse call processing from the local library management unit to the host unit, while making full use of the two-way interaction capabilities of the library interface to ensure that the call task is accurate and efficient.

[0071] If it is a transfer task type, the forward task processing of the conference data, the reverse call processing based on the stored conference data, and the forward transfer processing are executed. The core of the transfer task is to realize the safe migration of data between the source node and the target node. Specifically, the upper unit sends a call request to the source local library management unit through the library interface, triggers the reverse call processing, and obtains the target data from the source node; after the data is uploaded to the upper unit, it is processed in combination with the task requirements (such as format conversion or sharding operation); then, the upper unit passes the processed data to the target local library management unit through the library interface, triggering the forward transfer processing. After the target node completes the storage, the task completion status is fed back through the library interface. Through the combination of forward task processing and reverse call processing, the system ensures the complete and safe migration of data between the source node and the target node.

[0072] Through the bidirectional data interaction design of the library interface, the forward processing (host unit to local library management unit) and reverse call (local library management unit to host unit) of the task are supported to achieve the flexibility and efficiency of data flow. Finally, the storage, call and transfer tasks can be completed quickly in a distributed environment, providing accurate and efficient technical support for conference data management.

[0073] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0074] The present application determines a distributed conference database and specifies the management requirements of conference data, wherein the conference database includes multiple conference databases and is marked with a unique library code; the management requirements are decomposed to determine the management requirements within the library for one-step training, and the coordination requirements between libraries are determined for two-step training, and a library management unit and a host unit are generated; the library management unit is called and delegated to the conference database for multiple rounds of internal competitive training based on local library characteristics to determine a local library management unit, wherein the local library management unit corresponds to the conference database one-to-one and is built into the conference database; based on the library interface, a connection is established between the local library management unit and the host unit to form a management model, wherein the library interface is marked with a library code; conference data management tasks are received, and task management is performed in combination with the management model, wherein task management includes sub-task allocation based on the host unit and multi-library parallel processing based on the local library management unit. The present invention solves the technical problems of inaccurate task allocation, low coordination efficiency and insufficient data processing capability in the prior art in distributed conference data management. Through distributed database design, a large management model is constructed using a unique library coding identifier and a library interface. A library management unit and a host unit are generated through one-step training and two-step training. Multiple rounds of internal competitive training are performed in combination with local library characteristics to optimize the local library management unit, realize accurate task allocation and multi-library parallel processing, and achieve the technical effect of improving the efficiency and coordination capability of conference data management.

[0075] Embodiment 2 is based on the same inventive concept as the conference data management system based on artificial intelligence in the above embodiment. Figure 2 As shown, the embodiment of the present application provides a conference data management method based on artificial intelligence, the method comprising:

[0076] Determine a distributed conference database and specify the management requirements of conference data, wherein the conference database includes multiple databases and is identified with a unique database code; decompose the management requirements, determine the management requirements within the database for one-step training, determine the coordination requirements between the databases for two-step training, and generate a database management unit and a host unit; call the database management unit and delegate it to the conference database for multiple rounds of internal competitive training based on local database characteristics to determine the local database management unit, wherein the local database management unit corresponds to the conference database one-to-one and is built into the conference database; establish a connection between the local database management unit and the host unit based on the library interface to form a large management model, wherein the library interface is marked with a library code; receive conference data management tasks, and perform task management in combination with the large management model, wherein task management includes subtask allocation based on the host unit and multi-library parallel processing based on the local database management unit.

[0077] Further, the management requirements within the library are determined to perform a training step, and the method further includes:

[0078] Based on the management requirements within the library, determine the standardized management logic, wherein the standardized management logic is universal to the library; call the data management record, perform record integration based on the standardized management logic, and determine the standardized sample; based on the standardized sample, perform sample-driven training until convergence to generate the library management unit.

[0079] Further, determining the need for inter-library coordination to conduct two-step training, the method further includes:

[0080] Set multiple management targets, perform coordinated training under the multiple management targets, and determine the first host device; perform random updates on the management targets, perform update management training under the multiple management targets on the first host device, and generate the host device unit.

[0081] Furthermore, multiple rounds of internal competitive training based on local library characteristics are performed, and the method further includes:

[0082] The library management unit is decentralized to a first conference database, and a first local library feature of the first conference database is called, where the first conference database is any conference database; an initial learning rate is determined for the first local library feature, wherein each local library feature corresponds to an initial learning rate; based on the initial learning rate, a round of training is performed on the library management unit to determine a round of management units; for the round of management units, an internal adjustment of the learning rate based on the local management standard and multiple rounds of iterative training are performed to generate a first local library management unit.

[0083] Further, the learning rate internal competition adjustment and multiple rounds of iterative training based on the local management standard are performed, and the method also includes:

[0084] The first-round management unit is verified, and a management evaluation is performed based on the local management standard to determine an evaluation matrix, wherein the evaluation matrix has local library features as matrix rows, and feature management coefficients and fusion management coefficients as digital columns, and the fusion management coefficient is the degree of fusion between the standardized management logic and the first local library features; the evaluation matrix is ​​used to perform internal competition adjustment between learning rates, and two rounds of training are performed to determine the second-round management unit; for the second-round management unit, management evaluation and iterative training are performed until the preset convergence conditions are met to generate the first local library management unit.

[0085] Furthermore, in combination with the management macro model, task management is performed, and the method further includes:

[0086] The conference data management task is uploaded to the upper device unit, task allocation based on library storage type and library code matching are performed to determine the management subtask; based on the library interface, the management subtask is sent to the target local library management unit, and the library management of the subtask is executed in parallel.

[0087] Furthermore, the method further comprises:

[0088] If it is a storage task type, execute forward task processing based on the conference data management task and the newly added conference data, wherein the host unit-local library management unit is the forward direction; if it is a calling task type, execute forward task processing of the conference data management task and reverse calling processing of the stored conference data, wherein the host unit is called through the library interface to perform reverse allocation and interface display, and the library interface can perform two-way data interaction; if it is a transfer task type, execute forward task processing of the conference data management task, reverse calling processing based on the stored conference data, and forward transfer processing based on the stored conference data.

[0089] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0090] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0091] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A conference data management system based on artificial intelligence, characterized in that: The system comprises: A management requirement determination module, the management requirement determination module is used to determine a distributed conference database and specify the management requirements of conference data, wherein the conference database includes multiple conference databases, each of which is identified by a unique database code; A first training module, the first training module is used to decompose the management requirements, determine the management requirements within the library for one-step training, determine the coordination requirements between libraries for two-step training, and generate a library management unit and a host unit; A second training module, the second training module is used to call the library management unit, and transfer it to the conference database to perform multiple rounds of internal competition training based on local library characteristics to determine the local library management unit, wherein the local library management unit corresponds to the conference database one by one and is built into the conference database; A management large model construction module, wherein the management large model construction module establishes a connection between the local library management unit and the host unit based on a library interface to form a management large model, wherein the library interface is marked with a library code; A task management module receives conference data management tasks and performs task management in combination with the management model, wherein task management includes subtask allocation based on a host unit and multi-library parallel processing based on a local library management unit.

2. The conference data management system based on artificial intelligence as claimed in claim 1, characterized in that: The first training module is used to: Based on the management requirements within the library, determine the standardized management logic, wherein the standardized management logic is universal to the library; Calling data management records, integrating the records based on the standardized management logic, and determining standardized samples; Based on the standardized samples, sample-driven training is performed until convergence to generate the library management unit.

3. The conference data management system based on artificial intelligence as claimed in claim 1, characterized in that: The first training module is used to: Setting multiple management goals, executing coordinated training under the multiple management goals, and determining the first upper controller; Randomly update the management target, execute update management training under the multiple management targets on the first host device, and generate the host device unit.

4. The conference data management system based on artificial intelligence as claimed in claim 2, characterized in that: The second training module is used to: Decentralize the library management unit to a first conference database, and call a first local library feature of the first conference database, where the first conference database is any conference database; Determine an initial learning rate for the first local library feature, wherein each local library feature corresponds to an initial learning rate; Based on the initial learning rate, performing a round of training on the library management unit to determine a round of management units; For the one-round management unit, internal competition adjustment of the learning rate and multiple rounds of iterative training are performed based on the local management standard to generate a first local library management unit.

5. The conference data management system based on artificial intelligence as claimed in claim 4, characterized in that: The second training module is used to: Verify the one-round management unit, perform management evaluation based on the local management standard, and determine an evaluation matrix, wherein the evaluation matrix has local library features as matrix rows, feature management coefficients and fusion management coefficients as digital columns, and the fusion management coefficient is the degree of fusion of the standardized management logic and the first local library features; Using the evaluation array, internal competition adjustment is performed between learning rates, and two rounds of training are performed to determine the two rounds of management units; For the second round management unit, the first local library management unit is generated through management evaluation and iterative training until a preset convergence condition is met.

6. The conference data management system based on artificial intelligence as claimed in claim 1, characterized in that: Task management module for: Upload the conference data management task to the upper device unit, perform task allocation based on library storage type and library code matching, and determine management subtasks; Based on the library interface, the management subtask is sent to the target local library management unit, and the library management of the subtask is executed in parallel.

7. The conference data management system based on artificial intelligence as claimed in claim 6, characterized in that: The system further comprises: If it is a storage task type, perform forward task processing based on the conference data management task and the newly added conference data, wherein the host unit-local library management unit is the forward direction; If it is a calling task type, the forward task processing of the conference data management task and the reverse calling processing of the storage conference data are executed, wherein the upper device unit is called through the library interface, reverse allocation is performed and the interface is displayed, and the library interface can perform two-way data interaction; If it is a transfer task type, the forward task processing of the conference data management task, the reverse call processing based on the stored conference data, and the forward transfer processing based on the stored conference data are executed.

8. A conference data management method based on artificial intelligence, characterized in that: The method is performed by an artificial intelligence-based conference data management system according to any one of claims 1 to 7, comprising: Determine a distributed conference database and specify the management requirements of conference data, wherein the conference database includes multiple databases and is identified by a unique database code; Decompose the management requirements, determine the management requirements within the library for one-step training, determine the coordination requirements between libraries for two-step training, and generate a library management unit and a host unit; Calling the library management unit, delegating it to the conference database to perform multiple rounds of internal competitive training based on local library characteristics, and determining a local library management unit, wherein the local library management unit corresponds to the conference database one by one and is built into the conference database; Based on the library interface, the connection between the local library management unit and the host unit is established to form a management model, wherein the library interface is marked with a library code; Receive conference data management tasks, and perform task management in combination with the management macro model, wherein the task management includes subtask allocation based on the host unit and multi-library parallel processing based on the local library management unit.

Images