Task execution method and device, electronic equipment and storage medium

Abstract

The present disclosure relates to a task execution method and device, electronic equipment and storage medium. The task type recognition and data type matching are performed on the to-be-processed task described in natural language, the target task type and the target data type corresponding to the to-be-processed task are determined, the to-be-processed task is split according to the determined result, a subtask structure diagram is generated, the execution tool corresponding to each subtask is obtained from a preset tool library, the parameterized calling is performed by using the execution tool corresponding to each subtask, the initial execution result of each subtask is generated and the regularized output control is performed, and the target execution result of each subtask is obtained. Therefore, the natural language interaction driven task execution method can be used to autonomously complete task analysis, tool matching, task execution and regularized output control, the task execution result can be quickly and accurately determined in a multi-source heterogeneous data and multi-agency collaborative scene, and the computer knowledge of relevant personnel is not required.

Description

Technical Field

[0001] This disclosure relates to the field of information processing technology, and in particular to a task execution method, apparatus, electronic device, and storage medium. Background Technology

[0002] With the rapid development of computer technology and the increasing complexity of information, using online task execution platforms can improve task processing efficiency and effectiveness. For example, in the field of new traditional Chinese medicine (TCM) drugs, the amount of TCM information is large and the data formats are diverse. Using online platforms to search for, place, and perform other related tasks for TCM drugs can improve the work efficiency and effectiveness of relevant personnel.

[0003] However, considering the multi-source heterogeneous data and multi-organizational collaboration issues involved in these scenarios, and the fact that personnel in these scenarios generally lack computer modeling skills, providing a task execution method that is convenient for personnel in these scenarios to improve task execution efficiency and effectiveness is a pressing technical problem that needs to be solved. Summary of the Invention

[0004] To address the aforementioned technical problems, this disclosure provides a task execution method, apparatus, electronic device, and storage medium.

[0005] Firstly, this disclosure provides a task execution method, including: For tasks described in natural language, task type identification and data type matching are performed to determine the target task type and target data type corresponding to the task to be processed. Based on the target task type and the target data type, the task to be processed is split into sub-tasks to generate a sub-task structure diagram containing multiple sub-tasks; From the preset tool library, obtain the execution tool corresponding to each subtask in the subtask structure diagram, and use the execution tool corresponding to each subtask to perform parameterized calls to generate the initial execution result of each subtask; The initial execution result of each subtask is subjected to rule-based output control to obtain the target execution result of each subtask.

[0006] Secondly, this disclosure provides a task execution apparatus, comprising: The first determining module is used to identify the task type and match the data type of the task to be processed, which is described in natural language, and to determine the target task type and target data type corresponding to the task to be processed. The first generation module is used to split the task to be processed based on the target task type and the target data type, and generate a subtask structure diagram containing multiple subtasks. The second generation module is used to obtain the execution tool corresponding to each subtask in the subtask structure diagram from the preset tool library, and use the execution tool corresponding to each subtask to perform parameterized calls to generate the initial execution result of each subtask; The second determining module is used to perform rule-based output control on the initial execution result of each subtask to obtain the target execution result of each subtask.

[0007] Thirdly, embodiments of this disclosure also provide an electronic device, the electronic device comprising: One or more processors; Storage device for storing one or more programs. When one or more programs are executed by one or more processors, the one or more processors implement the methods provided in the first aspect.

[0008] Fourthly, embodiments of this disclosure also provide a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method provided in the first aspect.

[0009] The technical solution provided in this disclosure has the following advantages compared with the prior art: This disclosure discloses a task execution method, apparatus, electronic device, and storage medium. First, it identifies the task type and matches the data type of the task to be processed, described in natural language, to determine the target task type and target data type. Then, based on the target task type and target data type, it breaks down the task to be processed, generating a subtask structure diagram containing multiple subtasks. Next, it retrieves the execution tool corresponding to each subtask in the subtask structure diagram from a preset tool library and uses the execution tool to perform parameterized calls, generating the initial execution result of each subtask. Finally, it performs rule-based output control on the initial execution result of each subtask to obtain the target execution result of each subtask. Therefore, this method can drive task execution with natural language interaction, autonomously completing task analysis, tool matching, task execution, and rule-based output control. It enables rapid and accurate determination of task execution results in scenarios involving multi-source heterogeneous data and multi-organizational collaboration, without relying on the computer knowledge of relevant personnel. Attached Figure Description

[0010] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0011] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A flowchart illustrating a task execution method provided in an embodiment of this disclosure; Figure 2 This is a schematic diagram of the structure of a task execution device provided in an embodiment of the present disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0013] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0014] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0015] In the field of new Chinese medicine drugs, the information on Chinese medicine drugs mainly includes information on medicinal materials / processed pieces / extracts / components / monomers, processing and process parameters, batch and supply chain traceability, efficacy and safety evaluation data, experimental data, omics, metabolomics, genomic data, etc., which are extremely large in volume and diverse in format.

[0016] In addition, personnel in the field of new Chinese medicine drugs generally lack computer modeling knowledge, which leads to problems such as experience in task selection and method path dependence, easy misuse, high threshold for script configuration, error location, parameter optimization, data governance and analysis, high cost of pre-cleaning and standardization, insufficient process traceability, and difficulty in reproducing results.

[0017] Meanwhile, new drug development data contains information that cannot be disclosed and sensitive information, which usually requires that the data not leave the domain when the task is processed. However, traditional practices often struggle to balance ease of use and compliance.

[0018] To provide a convenient task execution method for relevant personnel in specific scenarios, thereby improving task execution efficiency and effectiveness, this embodiment provides a task execution method. The following is a combination of... Figure 1The task execution method provided in this disclosure is described below. In this disclosure, the task execution method can be executed by an electronic device or a server. The electronic device may include a desktop computer, laptop, tablet, and other smart devices. The server may include a cloud server or a server cluster. Specifically, this embodiment uses a smart agent within an electronic device as an example to perform task execution.

[0019] Figure 1 A flowchart illustrating a task execution method provided in an embodiment of this disclosure is shown.

[0020] As shown in Figure 1, the task execution method may include the following steps.

[0021] S110. For the task to be processed described in natural language, perform task type identification and data type matching to determine the target task type and target data type corresponding to the task to be processed.

[0022] In this embodiment, the user can drive the intelligent agent through natural language interaction to autonomously execute tasks. Specifically, the electronic device obtains the target and constraints described in natural language through the natural language interaction function of the intelligent agent, and uses multi-identity authentication and role permission functions, as well as multi-data type access and resource directory functions, to complete semantic analysis, identity authentication, and recognition tasks, thereby determining the target task type and target data type.

[0023] Among them, the tasks to be processed refer to the tasks that need to be processed in the relevant application scenarios.

[0024] For example, in the field of new traditional Chinese medicine drugs, the task to be processed can be an important classification task or an important query task.

[0025] For example, in the field of education, the task to be processed could be a student information retrieval task or a teacher information retrieval task.

[0026] Optionally, the target task type includes data governance type, data analysis type, and a combination of data governance type and data analysis type.

[0027] It should be noted that electronic devices can switch between different task types, forming an end-to-end closed loop.

[0028] Data governance tasks cover the entire chain from data access to quality consolidation. Specifically, they include one or more combinations of the following: data access and cataloging, terminology and coding alignment, data quality assessment, anonymization and aggregation strategy configuration, data version management, and consolidation of quality rules into usable tools.

[0029] Among them, data analysis tasks cover the entire process from exploratory analysis to report output, specifically including one or more combinations of the following: descriptive statistics, difference and comparison analysis, correlation and association mining, modeling training and evaluation within the scope of tool whitelists and compliance constraints, quality consistency and inter-batch difference analysis, exploration and validation, data result value assessment, result visualization and automatic report generation, etc.

[0030] Optionally, the target data types include structured data, semi-structured data, text data, experimental instrument and spectral feature data, omics data, time series and longitudinal data.

[0031] In some embodiments, the specific implementation method of S110 includes: inputting the task to be processed into a preset task classifier for task type identification to obtain the target task type; performing correlation analysis on preset multiple data types and the task to be processed to obtain the target data type.

[0032] Optionally, the methods for determining the target task type include:

[0033] in, For multi-class classifiers; A vectorized representation of the task to be processed; , These are learnable parameters.

[0034] Optionally, the methods for determining the target data type include:

[0035] in, The Sigmoid function maps linear scores to... interval; For data types The feature weight vector, ={structured data, semi-structured data, text data, experimental instrument and spectral feature data, omics data, time series and longitudinal data}; For data types The activation threshold. The specific principle of the method for determining the target data type is to determine the semantic relevance between the vectorized representation corresponding to the task to be processed and different data types, and only output the data types whose semantic relevance exceeds the threshold as the target data type, thereby driving the corresponding data access and processing adapter.

[0036] S120. Based on the target task type and target data type, the task to be processed is split into subtasks to generate a subtask structure diagram containing multiple subtasks.

[0037] When the task to be processed is complex, in order to improve the efficiency of task processing, the task to be processed can be broken down into multiple subtasks, and the task to be processed can be achieved by processing the subtasks.

[0038] The subtask structure graph can be a directed acyclic graph formed by subtasks as nodes and the dependencies between subtasks as edges.

[0039] In some embodiments, the specific implementation method of S120 includes: obtaining the target adapter corresponding to the task to be processed from the preset adapters according to the target task type and the target data type; using the target adapter, with the minimum total execution cost as the constraint, splitting the task to be processed according to the execution logic between multiple subtasks to generate a subtask structure diagram.

[0040] Specifically, the electronic device is pre-configured with appropriate adapters based on different task types and data types. These adapters can handle the processing logic of tasks of different types and data types. Therefore, after determining the target data type, the corresponding adapter is directly invoked as the target adapter for the task to be processed. The target adapter is then used to break down the task into multiple subtasks, and the total execution cost of the multiple subtasks is calculated. The subtask structure diagram is generated with the minimum total execution cost as the constraint.

[0041] Optionally, the method for determining the subtask structure graph includes:

[0042] in, It is the coverage of user needs; A collection of user needs.

[0043] In this way, under the constraint of meeting the coverage of user needs, the combination of subtasks with the lowest total execution cost is selected, and a subtask structure diagram is generated based on the dependencies between the subtasks.

[0044] In some embodiments, after executing S120, the method further includes: determining the data size, operator complexity, and predicted iteration number of the target task according to the subtask structure diagram; performing a weighted summation of the data size, operator complexity, and the predicted iteration number to obtain the task complexity of the task to be processed; if the task complexity is greater than a preset complexity threshold, then performing a breakpoint resume operation on the task to be processed in the target execution environment.

[0045] The target execution environment can be understood as a data space sandbox. This environment provides a completely isolated execution environment for subtasks, prohibiting external network connections, restricting file system access, and only allowing results to be written through a controlled output area, thereby ensuring that data does not leave the domain at the technical level.

[0046] Optionally, the target execution environment includes one or more of the following constraints: constraints on data access performed through gateway functions, constraints on computation performed in the target execution environment, constraints on controlled output area writes, and constraints on resource budgets.

[0047] The data size is used to measure the impact of the data volume on the computational load. Specifically, the data size is obtained by logarithmically normalizing the data volume of the subtask structure graph.

[0048] Operator complexity is used to measure the computational complexity of the combined operations invoked by the task. Specifically, operator complexity is determined by taking the average of the time complexity records of all invoked operators.

[0049] The preset iteration count is used to measure the estimated number of loop or recursive iterations during execution. Specifically, it can be based on the system upper limit. The ratio is normalized and mapped to the interval [0, 3] to determine the preset number of iterations.

[0050] Optionally, the methods for determining the task complexity of the task to be processed include:

[0051] in, It refers to the scale of the data; It is operator complexity; It predicts the number of iterations; , , These are weights that are greater than or equal to 0. , , The relative contribution of each dimension to task complexity under different scenarios can be determined by regression fitting of historical task data, and... .

[0052] Understandably, higher task complexity indicates a heavier workload. If the task complexity exceeds a preset complexity threshold, it is determined to be a high-complexity task, requiring it to enter a long-cycle asynchronous execution mode to perform breakpoint resume operations.

[0053] The specific implementation method for resuming the task under processing in the target execution environment includes: dividing the task under processing into multiple task stages in the target execution environment and determining the checkpoint of each task stage; if execution failure information or interruption information is detected at the checkpoint of the target task stage, then obtaining the task stage to be recovered that is closest to the target task stage from the multiple task stages; starting from the task stage to be recovered, performing the resuming operation on the task under processing.

[0054] Optionally, checkpoints for each task phase can be represented as follows:

[0055] in, This represents the number of task stages for the tasks yet to be processed. It is the first The execution function of the phase; These are the corresponding parameters. Specifically, upon completion of each stage, intermediate states (including data snapshots, model weights, and execution locations) can be persisted as checkpoints. .

[0056] To ensure that execution can continue from the previous completed checkpoint after any stage termination, after an execution failure or interruption message is detected at the checkpoint of the target task stage, the task stage to be recovered that is closest to the target task stage is selected from multiple task stages, i.e., the nearest checkpoint is obtained. Then, starting from the task stage to be recovered, the task to be processed is executed to resume from the breakpoint, thereby avoiding the waste of resources caused by starting from the beginning.

[0057] In this way, by using an isolated sandbox within the domain to support long-cycle asynchronous tasks, breakpoint resume, and resource quotas, the resource consumption of highly complex tasks can be reduced while controlling output risks.

[0058] S130. Obtain the execution tool corresponding to each subtask in the subtask structure diagram from the preset tool library, and use the execution tool corresponding to each subtask to perform parameterized calls to generate the initial execution result of each subtask.

[0059] In generating the subtask structure diagram, the electronic device needs to configure a corresponding execution tool for each subtask in the subtask structure diagram, and then process the subtask by parameterizing the execution tool, thereby obtaining the initial execution result of each subtask.

[0060] In some embodiments, the specific implementation method of "obtaining the execution tool corresponding to each subtask in the subtask structure diagram from the preset tool library" in S130 includes: obtaining the capability matching degree and data matching degree between each subtask and each tool in the preset tool library, and obtaining the compliance data and cost data of each tool in the preset tool library; calculating the matching data between each subtask and multiple tools based on the capability matching degree, data matching degree, compliance data and cost data; and obtaining the execution tool corresponding to each subtask from the preset tool library according to the matching data between each subtask and multiple tools.

[0061] Capability matching degree is used to measure the semantic matching degree between tool functions and subtask requirements. Specifically, the capability matching degree is determined by calculating the cosine similarity between the tool capability description embedding and the subtask requirement embedding.

[0062] Data matching degree is used to measure the tool's compatibility with the current data type. Specifically, it determines the data matching degree by obtaining the coverage ratio between the data types supported by the tool and the data types of the current task.

[0063] Compliance data is a prerequisite for excluding non-compliant tools. When a tool meets the current policy, the compliance data is 1, and the tool is retained; when a tool does not meet the current policy, the compliance data is 0, and the tool is directly excluded.

[0064] Cost data is used to measure the resource consumption of tools, and tools with lower resource consumption are selected while ensuring functionality.

[0065] Optionally, the method for determining the matching data between each subtask and multiple tools includes:

[0066] in, It is a tool sub-tasks The degree of compatibility between their capabilities; It is a subtask data types With tools The degree of data matching between them; In the current strategy Time tools Compliance data; It is a tool Cost data; , , , All weights are greater than 0, and + + + =1.

[0067] The specific implementation method for obtaining the execution tool corresponding to each subtask includes: obtaining the matching data with the largest value from the matching data between each subtask and multiple tools; if the matching data with the largest value is greater than or equal to a preset matching threshold, then the tool corresponding to the matching data with the largest value is used as the execution tool corresponding to each subtask; if the matching data with the largest value is less than the preset matching threshold, then the execution tool self-growth operation is performed to obtain the execution tool corresponding to each subtask.

[0068] Optionally, methods for determining the matching data with the largest value include:

[0069] Understandably, if the largest matching data value is greater than or equal to the preset matching threshold, it means that the tool corresponding to the largest matching data value can perform the corresponding sub-task well. In this case, the tool corresponding to the largest matching data value will be used as the execution tool for each sub-task. If the largest matching data value is less than the preset matching threshold, it means that the tool corresponding to the largest matching data value lacks the ability to perform the task. In this case, the execution tool for the sub-task can be regenerated according to the tool self-growth strategy.

[0070] The specific implementation method of the self-growth operation of the execution tool to obtain the execution tool corresponding to each subtask includes: in the target execution environment, based on each subtask and a preset tool library, the self-growth operation of the execution tool is performed to generate candidate tools, wherein the target execution environment is an execution environment isolated from the outside; the candidate tools are subjected to security and compliance verification to obtain the candidate tools that pass the verification; the candidate tools that pass the verification are tested and verified, and the candidate tools that pass the test and verification are used as the execution tools corresponding to each subtask.

[0071] Alternatively, constraints on data access executed through gateway functions can be represented as follows:

[0072] Alternatively, the constraints for performing computations in the target execution environment can be represented as follows:

[0073] Optionally, the constraints on controlled output area writes can be represented as follows:

[0074] Alternatively, resource budget constraints can be expressed as follows:

[0075] in, , It is the CPU resource budget. It is the memory resource budget. It is the GPU resource budget. This is a time resource budget. This constraint requires task forecasting resources. No more than Execution is only allowed when certain conditions are met; otherwise, a degradation strategy is triggered, such as sampling, dimensionality reduction, or reducing iterations.

[0076] Optionally, security and compliance verification may include external connection verification, privilege escalation verification, write path verification, and plaintext disclosure verification. Specifically, after the candidate tools pass external connection verification, privilege escalation verification, write path verification, and plaintext disclosure verification, the verified candidate tools are obtained. Then, the verified candidate tools are tested and verified on synthetic data or controlled sample sets. The verified candidate tools are used as the execution tools corresponding to each subtask, and the execution tools are registered in the tool library for easy subsequent use.

[0077] In some embodiments, the specific implementation method of "using the execution tool corresponding to each subtask to perform parameterized calls and generate the initial execution result of each subtask" in S130 includes: obtaining the tool parameters of the execution tool corresponding to each subtask; constraining the tool parameters of the execution tool corresponding to each subtask to obtain the constrained tool parameters corresponding to each subtask; and calling the execution tool corresponding to each subtask to perform parameterized calls according to the constrained tool parameters corresponding to each subtask to generate the initial execution result of each subtask.

[0078] Optionally, the method for determining the constrained tool parameters corresponding to each subtask includes:

[0079] in, These are the tool parameters for the execution tool corresponding to each subtask; It is a comprehensive loss value for the target's overall stability, resources, etc. It is the constraint of the compliance strategy set on the parameter space.

[0080] By employing the above methods, including autonomous task decomposition and dynamic tool selection, tasks are automatically broken down into multiple subtasks, which are then dynamically matched and parameterized in the tool library, lowering the scripting threshold. When a tool lacks the required capabilities, candidate tools are generated within the data space sandbox based on a tool self-growth strategy and a controlled registration strategy. After security and compliance verification and testing, the candidate tools are used to obtain the execution tools, which are then registered in the tool library, enabling continuous expansion of tool capabilities.

[0081] Furthermore, by constraining the tool parameters of the execution tool corresponding to each subtask to ensure that the parameters are within the feasible domain that meets the compliance constraints, and with the goal of minimizing task execution loss, the constrained tool parameters corresponding to each subtask are determined, and the execution tool is called according to the tool parameters to achieve parameterized calls, thereby obtaining the initial execution result of each subtask.

[0082] S140. Perform regularized output control on the initial execution results of each subtask to obtain the target execution result of each subtask.

[0083] To reduce compliance risks and management costs, it is also necessary to implement rule-based output control for the initial execution results of each sub-task, so as to achieve internalized compliance and trustworthiness rather than post-event review.

[0084] In some embodiments, the specific implementation method of S140 includes: performing parallel rule-based output control on the initial execution results of each subtask, including rule-based output control on field types, rule-based output control on identifiable risk types, and rule-based output control on desensitized aggregation types, to obtain the target execution result of each subtask.

[0085] Optionally, the regularized output control of field classes can be represented in the following way:

[0086] in, It is a set of request plans; This is a field that is allowed by permissions.

[0087] Optionally, rule-based output control that identifies risk classes can be represented as follows:

[0088] in, It is an output grouping unit The count; It is a threshold.

[0089] Alternatively, the regularized output control of the desensitized aggregation class can be represented by the following method:

[0090] in, This means executing from right to left. This refers to performing generalization on units below a threshold to reduce identifiable risks; This refers to suppressing identifiable cells in small samples; This refers to masking and replacing personal identity and sensitive fields.

[0091] In this way, by controlling the output of subtasks in a rule-based manner, the safety principles of ensuring that data does not leave the domain and that results are output under control are met.

[0092] This disclosure discloses a task execution method. First, it identifies the task type and matches the data type of the task to be processed, described in natural language, to determine the target task type and target data type. Then, based on the target task type and target data type, it breaks down the task to be processed, generating a subtask structure diagram containing multiple subtasks. Next, it retrieves the execution tool corresponding to each subtask in the subtask structure diagram from a preset tool library, and uses the execution tool corresponding to each subtask to perform parameterized calls, generating the initial execution result of each subtask. Finally, it performs rule-based output control on the initial execution result of each subtask to obtain the target execution result of each subtask. Therefore, this method can drive task execution with natural language interaction, autonomously completing task analysis, tool matching, task execution, and rule-based output control. It enables rapid and accurate determination of task execution results in scenarios involving multi-source heterogeneous data and multi-institutional collaboration, and can complete task execution without relying on the computer knowledge of relevant personnel.

[0093] In some embodiments, after S140, the method further includes: generating record information and basis information corresponding to the target execution result of each subtask, and storing the record information, basis information and the target execution result of each subtask in association.

[0094] Among these, the recorded information can be represented in the form of audit records.

[0095] The supporting information can be an evidence package, specifically obtained by hashing key information within the recorded information. Optionally, the key information may include code information, image information, parameters, plans, result summaries, etc.

[0096] In this way, by combining the audit chain and evidence package, a reliable control system for full-chain evidence storage, traceability, and operational auditing can be established to ensure the security, controllability, and traceability of the entire data usage process.

[0097] Through the above methods, a complete process is formed from data collection to analysis and results distribution, achieving standardized, traceable, and easily reproducible task processes.

[0098] This disclosure also provides a task execution apparatus for implementing the above-described task execution method, which will be described below in conjunction with... Figure 2 The following explanation is provided. In this embodiment, the task execution method can be executed by an electronic device or a server. The electronic device may include a desktop computer, laptop, tablet, and other smart devices. The server may include a cloud server or a server cluster. This embodiment uses an electronic device as the execution subject for specific explanation.

[0099] Figure 2 A schematic diagram of the structure of a task execution device provided in an embodiment of this disclosure is shown.

[0100] like Figure 2 As shown, the task execution device 200 may include: The first determining module 210 is used to identify the task type and match the data type of the task to be processed, which is described in natural language, and to determine the target task type and target data type corresponding to the task to be processed. The first generation module 220 is used to split the task to be processed based on the target task type and the target data type, and generate a subtask structure diagram containing multiple subtasks. The second generation module 230 is used to obtain the execution tool corresponding to each subtask in the subtask structure diagram from the preset tool library, and use the execution tool corresponding to each subtask to perform parameterized calls to generate the initial execution result of each subtask; The second determining module 240 is used to perform rule-based output control on the initial execution result of each subtask to obtain the target execution result of each subtask.

[0101] An embodiment of this disclosure discloses a task execution device. First, it identifies the task type and matches the data type of the task to be processed, described in natural language, to determine the target task type and target data type corresponding to the task to be processed. Then, based on the target task type and target data type, it breaks down the task to be processed to generate a subtask structure diagram containing multiple subtasks. Next, it retrieves the execution tool corresponding to each subtask in the subtask structure diagram from a preset tool library, and uses the execution tool corresponding to each subtask to perform parameterized calls to generate the initial execution result of each subtask. Finally, it performs rule-based output control on the initial execution result of each subtask to obtain the target execution result of each subtask. Thus, it can drive the task execution method with natural language interaction, autonomously complete task analysis, tool matching, task execution, and rule-based output control, and achieve rapid and accurate determination of task execution results in scenarios with multi-source heterogeneous data and multi-institutional collaboration, without relying on the computer knowledge of relevant personnel to complete task execution.

[0102] In some embodiments of this disclosure, the first determining module 210 includes: The task identification unit is used to input the task to be processed into a preset task classifier to identify the task type and obtain the target task type. The correlation analysis unit is used to perform correlation analysis between multiple preset data types and the task to be processed to obtain the target data type.

[0103] In some embodiments of this disclosure, the first generation module 220 includes: The first acquisition unit is used to acquire the target adapter corresponding to the task to be processed from a preset set of adapters according to the target task type and the target data type. The first generation unit is used to utilize the target adapter, with the minimum total execution cost as a constraint, to split the task to be processed according to the execution logic between the multiple subtasks, and generate the subtask structure diagram.

[0104] In some embodiments of this disclosure, the second generation module 230 includes: The second acquisition unit is used to acquire the capability matching degree and data matching degree between each of the sub-tasks and each tool in the preset tool library, and to acquire the compliance data and cost data of each tool in the preset tool library; The first calculation unit is used to calculate the matching data between each subtask and multiple tools based on the capability matching degree, the data matching degree, the compliance data, and the cost data; The third acquisition unit is used to acquire the execution tool corresponding to each subtask from the preset tool library based on the matching data between each subtask and multiple tools.

[0105] In some embodiments of this disclosure, the third acquisition unit includes: The first acquisition subunit is used to acquire the matching data with the largest value from the matching data between each subtask and multiple tools; The first determining subunit is used to select the tool corresponding to the largest matching data as the execution tool for each subtask if the largest matching data is greater than or equal to a preset matching threshold. The second acquisition subunit is used to perform a tool self-growth operation to acquire the execution tool corresponding to each subtask if the largest matching data is less than a preset matching threshold.

[0106] In some embodiments of this disclosure, the second acquisition subunit is specifically used for: In the target execution environment, based on each subtask and the preset tool library, a tool self-growth operation is performed to generate candidate tools, wherein the target execution environment is an execution environment isolated from the outside. The candidate tools are subjected to security and compliance verification to obtain the candidate tools that pass the verification; The candidate tools that pass the verification are tested and verified, and the candidate tools that pass the test and verification are used as the execution tools corresponding to each subtask.

[0107] In some embodiments of this disclosure, the target execution environment includes one or more of the following constraints: Data access is subject to constraints imposed by gateway functions, constraints imposed by computations performed in the target execution environment, constraints imposed by controlled output area writes, and constraints imposed by resource budgets.

[0108] In some embodiments of this disclosure, the second generation module 230 includes: The fourth acquisition unit is used to acquire the tool parameters of the execution tool corresponding to each subtask; The first determining unit is used to constrain the tool parameters of the execution tool corresponding to each subtask, and obtain the constrained tool parameters corresponding to each subtask. The second generation unit is used to call the execution tool corresponding to each subtask to perform parameterized calls according to the constrained tool parameters corresponding to each subtask, and generate the initial execution result of each subtask.

[0109] In some embodiments of this disclosure, the second determining module 240 includes: The second determining unit is used to perform parallel output control on the initial execution results of each subtask, including field-based rule-based output control, identifiable risk-based rule-based output control, and desensitized aggregation-based rule-based output control, to obtain the target execution result of each subtask.

[0110] In some embodiments of this disclosure, the device further includes: The third determining module is used to determine the data size, operator complexity, and predicted iteration number of the task to be processed based on the subtask structure diagram. The fourth determining module is used to perform a weighted summation of the data size, the operator complexity, and the number of prediction iterations to obtain the task complexity of the task to be processed. The breakpoint resume execution module is used to perform a breakpoint resume operation on the task to be processed in the target execution environment if the complexity of the task is greater than a preset complexity threshold.

[0111] In some embodiments of this disclosure, the breakpoint resume execution module includes: The third determining unit is used to divide the task to be processed into multiple task stages in the target execution environment and determine the checkpoint of each task stage. The fifth acquisition unit is used to acquire the task phase to be recovered that is closest to the target task phase from the plurality of task phases if execution failure information or interruption information is detected at the checkpoint of the target task phase. The breakpoint resume execution unit is used to perform breakpoint resume operation on the task to be processed starting from the task to be recovered stage.

[0112] In some embodiments of this disclosure, the device further includes: The associated storage module is used to generate record information and basis information corresponding to the target execution result of each subtask, and to associate and store the record information, the basis information and the target execution result of each subtask.

[0113] It should be noted that, Figure 2 The task execution device 200 shown can perform Figure 1 The various steps in the method embodiment shown are implemented. Figure 1 The processes and effects in the method embodiments shown are not described in detail here.

[0114] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure is shown.

[0115] like Figure 3 As shown, the electronic device may include a processor 301 and a memory 302 storing computer program instructions.

[0116] Specifically, the processor 301 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0117] Memory 302 may include a large-capacity storage for advertising or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 302 may include removable or non-removable (or fixed) media. Where appropriate, memory 302 may be internal or external to the integrated gateway device. In a particular embodiment, memory 302 is non-volatile solid-state memory. In a particular embodiment, memory 302 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (Electrically Programmable ROM, EPROM), an electrically erasable programmable PROM (EEPROM), an electrically alterable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0118] The processor 301 acquires and executes computer program instructions stored in the memory 302 to perform the steps of the task execution method provided in the embodiments of this disclosure.

[0119] In one example, the electronic device may also include a transceiver 303 and a bus 304. Wherein, as... Figure 3 As shown, the processor 301, memory 302 and transceiver 303 are connected via bus 304 and communicate with each other.

[0120] Bus 304 may include hardware, software, or both. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 304 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0121] The following are embodiments of a computer-readable storage medium provided in this disclosure. This computer-readable storage medium and the task execution methods of the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the computer-readable storage medium, please refer to the embodiments of the above task execution methods.

[0122] This embodiment provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a task execution method, including: For tasks described in natural language, task type identification and data type matching are performed to determine the target task type and target data type corresponding to the task to be processed. Based on the target task type and the target data type, the task to be processed is split into sub-tasks to generate a sub-task structure diagram containing multiple sub-tasks; From the preset tool library, obtain the execution tool corresponding to each subtask in the subtask structure diagram, and use the execution tool corresponding to each subtask to perform parameterized calls to generate the initial execution result of each subtask; The initial execution result of each subtask is subjected to rule-based output control to obtain the target execution result of each subtask.

[0123] Of course, the computer-executable instructions provided in the embodiments of this disclosure are not limited to the above-described method operations, but can also perform related operations in the information push method provided in any embodiment of this disclosure.

[0124] Based on the above description of the implementation methods, those skilled in the art can clearly understand that this disclosure can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer cloud platform (which may be a personal computer, a server, or a network cloud platform, etc.) to execute the information push method provided in the various embodiments of this disclosure.

[0125] Note that the above description is merely a preferred embodiment and the technical principles employed in this disclosure. Those skilled in the art will understand that this disclosure is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this disclosure. Therefore, although this disclosure has been described in detail through the above embodiments, it is not limited to the above embodiments. Many other equivalent embodiments may be included without departing from the concept of this disclosure, and the scope of this disclosure is determined by the scope of the appended claims.

Claims

1. A task execution method, characterized in that, include: For tasks described in natural language, task type identification and data type matching are performed to determine the target task type and target data type corresponding to the task to be processed. Based on the target task type and the target data type, the task to be processed is split into sub-tasks to generate a sub-task structure diagram containing multiple sub-tasks; From the preset tool library, obtain the execution tool corresponding to each subtask in the subtask structure diagram, and use the execution tool corresponding to each subtask to perform parameterized calls to generate the initial execution result of each subtask; The initial execution result of each subtask is subjected to rule-based output control to obtain the target execution result of each subtask.

2. The method according to claim 1, characterized in that, The step of identifying the task type and matching the data type of the task to be processed, which is described in natural language, to determine the target task type and target data type corresponding to the task to be processed includes: The task to be processed is input into a preset task classifier for task type identification to obtain the target task type; A correlation analysis is performed between the preset data types and the task to be processed to obtain the target data type.

3. The method according to claim 1, characterized in that, The step of splitting the task to be processed based on the target task type and the target data type to generate a subtask structure diagram containing multiple subtasks includes: Based on the target task type and the target data type, obtain the target adapter corresponding to the task to be processed from the preset adapters; Using the target adapter, with the minimum total execution cost as a constraint, the task to be processed is split according to the execution logic between the multiple subtasks to generate the subtask structure diagram.

4. The method according to claim 1, characterized in that, The step of obtaining the execution tool corresponding to each subtask in the subtask structure diagram from a preset tool library includes: Obtain the capability matching degree and data matching degree between each of the subtasks and each tool in the preset tool library, and obtain the compliance data and cost data of each tool in the preset tool library; Based on the capability matching degree, the data matching degree, the compliance data, and the cost data, the matching data between each subtask and multiple tools is calculated; Based on the matching data between each subtask and multiple tools, the execution tool corresponding to each subtask is obtained from the preset tool library.

5. The method according to claim 4, characterized in that, The step of obtaining the execution tool corresponding to each subtask from the preset tool library based on the matching data between each subtask and multiple tools includes: From the matching data between each subtask and multiple tools, obtain the matching data with the largest value; If the largest matching data is greater than or equal to the preset matching threshold, then the tool corresponding to the largest matching data will be used as the execution tool for each subtask. If the largest matching data is less than the preset matching threshold, the tool will perform a self-growth operation to obtain the execution tool corresponding to each subtask.

6. The method according to claim 5, characterized in that, The execution tool self-growth operation obtains the execution tool corresponding to each subtask, including: In the target execution environment, based on each subtask and the preset tool library, a tool self-growth operation is performed to generate candidate tools, wherein the target execution environment is an execution environment isolated from the outside. The candidate tools are subjected to security and compliance verification to obtain the candidate tools that pass the verification; The candidate tools that pass the verification are tested and verified, and the candidate tools that pass the test and verification are used as the execution tools corresponding to each subtask.

7. The method according to claim 6, characterized in that, The target execution environment includes one or more of the following constraints: Data access is subject to constraints imposed by gateway functions, constraints imposed by computations performed in the target execution environment, constraints imposed by controlled output area writes, and constraints imposed by resource budgets.

8. The method according to claim 1, characterized in that, The step of using the execution tool corresponding to each subtask to perform parameterized calls and generate the initial execution result of each subtask includes: Obtain the tool parameters of the execution tool corresponding to each subtask; Constrain the tool parameters of the execution tool corresponding to each subtask to obtain the constrained tool parameters for each subtask. According to the constrained tool parameters corresponding to each subtask, the execution tool corresponding to each subtask is called to perform parameterized calls, and the initial execution result of each subtask is generated.

9. The method according to claim 1, characterized in that, The step of performing rule-based output control on the initial execution results of each subtask to obtain the target execution result of each subtask includes: The initial execution results of each subtask are subjected to parallel rule-based output control for field types, rule-based output control for identifiable risk types, and rule-based output control for de-identification and aggregation types to obtain the target execution result of each subtask.

10. The method according to claim 1, characterized in that, After splitting the task to be processed based on the target task type and the target data type to generate a subtask structure diagram containing multiple subtasks, the method further includes: Based on the subtask structure diagram, determine the data size, operator complexity, and prediction iteration number of the task to be processed; The task complexity of the task to be processed is obtained by weighted summing of the data size, the operator complexity, and the number of prediction iterations. If the complexity of the task exceeds a preset complexity threshold, then the task to be processed will be executed with a breakpoint resume operation in the target execution environment.

11. The method according to claim 10, characterized in that, The step of performing a breakpoint resume operation on the task to be processed in the target execution environment includes: In the target execution environment, the task to be processed is divided into multiple task stages, and a checkpoint is determined for each task stage. If an execution failure or interruption message is detected at a checkpoint in the target task stage, then the task stage to be recovered that is closest to the target task stage is obtained from the plurality of task stages. Starting from the task to be recovered stage, the task to be processed is executed with a breakpoint resume operation.

12. The method according to claim 1, characterized in that, After performing rule-based output control on the initial execution results of each subtask to obtain the target execution result of each subtask, the method further includes: Generate record information and basis information corresponding to the target execution result of each subtask, and store the record information, basis information and the target execution result of each subtask in association.

13. A task execution device, characterized in that, include: The first determining module is used to identify the task type and match the data type of the task to be processed, which is described in natural language, and to determine the target task type and target data type corresponding to the task to be processed. The first generation module is used to split the task to be processed based on the target task type and the target data type, and generate a subtask structure diagram containing multiple subtasks. The second generation module is used to obtain the execution tool corresponding to each subtask in the subtask structure diagram from the preset tool library, and use the execution tool corresponding to each subtask to perform parameterized calls to generate the initial execution result of each subtask; The second determining module is used to perform rule-based output control on the initial execution result of each subtask to obtain the target execution result of each subtask.

14. An electronic device, characterized in that, include: processor; Memory, used to store executable instructions; The processor is configured to retrieve the executable instructions from the memory and execute the executable instructions to implement the method of any one of claims 1-12.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that, The storage medium stores a computer program that, when executed by a processor, causes the processor to implement the method described in any one of claims 1-12.

Images