A one-click scoring task distribution method and system for an ETP platform based on a workflow engine

Abstract

This invention relates to the field of intelligent scheduling technology and proposes a method and system for one-click scoring task distribution on an ETP platform based on a workflow engine. The method includes: allocating workflow instances to the scoring task set in the ETP platform according to the workflow engine; activating each branch node sequentially according to the topology order; and passing the scoring task group of each branch node into the input queue of the currently activated branch node; for the currently activated branch node, calling the expert resource library to dynamically match candidate experts for each scoring task in the branch node's input queue, and writing the distribution record into a distributed log storage; listening for completion events in the scoring task distribution workflow instances; and when all branch nodes report task completion, the aggregation node of the workflow engine collects the scoring results returned by each branch node, summarizes and generates a complete scoring report, and pushes the scoring report to the front-end interface. This invention can improve the efficiency of one-click scoring task distribution on the ETP platform.

Description

Technical Field

[0001] This invention relates to the field of intelligent scheduling technology, and in particular to a one-click scoring task distribution method and system for an ETP platform based on a workflow engine. Background Technology

[0002] Currently, most ETP platform scoring tasks are assigned manually and processed offline, lacking a standardized process scheduling mechanism. They cannot be reasonably arranged based on the inherent dependencies between tasks, making it difficult to achieve topologically ordered execution and concurrent control of task branches. Task assignment is highly random, the flow logic is chaotic, the overall task distribution efficiency is low, and the cost of manual intervention is high. They cannot meet the platform's need for rapid initiation and unified scheduling of large batches of scoring tasks.

[0003] The existing platform's scoring task allocation mode can only achieve basic random assignment of personnel, lacking multi-dimensional intelligent expert matching and screening logic. It fails to comprehensively optimize based on task attributes, expert professional suitability, historical scoring performance, and load status, which easily leads to problems such as mismatch of expert expertise, unbalanced task load, and poor consistency of scoring standards. At the same time, the task distribution process lacks a complete distributed log retention and full-process event monitoring mechanism, making it difficult to trace the task execution status in real time, and abnormal tasks cannot be detected and handled in a timely manner. The automation level of scoring result integration and summarization is low, and the timeliness and completeness of the final scoring report cannot be guaranteed. Therefore, how to improve the efficiency of one-click scoring task distribution on the ETP platform has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a one-click scoring task distribution method and system for an ETP platform based on a workflow engine, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a one-click scoring task distribution method for an ETP platform based on a workflow engine, comprising: S1. Based on the workflow engine, assign workflow instances to the scoring task set in the ETP platform, activate each branch node in sequence according to the topology, and pass the scoring task group corresponding to each branch node into the input queue of the currently activated branch node. S2. For the currently active branch node, call the expert resource library of the ETP platform to dynamically match candidate experts for each scoring task in the input queue of the branch node, and write the distribution record into the distributed log storage of the ETP platform. S3. Listen for the completion events of each branch node in the scoring task distribution workflow instance. When all branch nodes report task completion, the workflow engine's aggregation node collects the scoring results returned by each branch node, summarizes and generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

[0006] In a preferred embodiment, the step of allocating workflow instances to the scoring task set in the ETP platform according to the workflow engine includes: The workflow engine parses the dependency tags between various scoring tasks in the scoring task set of the ETP platform. If there are multiple scoring tasks with dependencies, the workflow engine will orchestrate them as serial branch nodes. If there is no dependency, the workflow engine will divide the scoring task into several parallel task groups according to the preset concurrency threshold, and each parallel task group corresponds to a parallel branch node. The workflow engine generates an independent sub-workflow execution context for each parallel branch node, and records the task identifier list of the scoring tasks contained in the parallel branch node in the sub-workflow execution context; The workflow engine connects serial and parallel branch nodes in topological order to generate a complete workflow definition file, and instantiates the workflow definition file to obtain a scoring task distribution workflow instance.

[0007] In a preferred embodiment, the step of activating each branch node sequentially according to the topological order and passing the scoring task group corresponding to each branch node into the input queue of the currently activated branch node includes: The workflow engine reads the node identifier of the starting branch node from the workflow instance and marks the starting branch node as active. For the currently active branch node, the workflow engine dynamically allocates the capacity of the input queue based on the number of tasks in the scoring task group corresponding to the branch node, and pushes each scoring task in the scoring task group into the input queue in turn. Once the currently active branch node has finished processing all scoring tasks in the input queue, the workflow engine queries the topological order of the workflow instance, determines the node identifier of the next branch node, and activates the next branch node, until all branch nodes are activated and processed.

[0008] In a preferred embodiment, the step of calling the expert resource library of the ETP platform for the currently active branch node to dynamically match candidate experts for each scoring task in the branch node's input queue includes: Analyze the category identifiers and task difficulty coefficients contained in the current scoring task; Based on the classification identifier, extract the first set of experts with corresponding expertise from the expert resource database; Based on the task difficulty coefficient, experts whose historical scoring accuracy is lower than the preset accuracy threshold are removed from the first expert set to obtain the second expert set; Obtain the number of tasks currently pending for each expert in the second expert set, and sort the second expert set in ascending order of the number of tasks currently pending. Select the expert at the top of the sorted set as the candidate expert for the current scoring task.

[0009] In a preferred embodiment, the step of dynamically matching candidate experts for each scoring task in the branch node input queue further includes: After selecting candidate experts, a second screening is conducted based on the experts' historical scoring preferences. Extract the distribution characteristics of the historical scores given by each candidate expert for past tasks with the same classification label as the current scoring task from the historical scoring records of the ETP platform. The scoring divergence of each candidate expert is calculated based on the distribution characteristics, and the scoring divergence is compared with the scoring consistency requirements of the current scoring task. Candidate experts whose scoring dispersion exceeds the scoring consistency requirement are removed. The remaining candidate experts are then sorted according to their historical collaboration frequency with the task initiator of the current scoring task, and the expert ranked first is selected as the final matching expert.

[0010] In a preferred embodiment, writing the distribution record into the distributed log storage of the ETP platform includes: The process record for matching candidate experts is integrated with the unique identifier of the scoring task, the identifier of the matched candidate expert, the distribution timestamp, and the node identifier of the branch node into a distribution record; The distribution record is encapsulated into a log entry, and the write interface of the distributed log storage is called to write the log entry to the currently active shard of the distributed log storage. After a log entry is successfully written, the distributed log storage returns a write confirmation message to the workflow engine and synchronizes the storage location index of the log entry to the audit log index table of the ETP platform.

[0011] In a preferred embodiment, the dispersion includes: ; In the formula, For experts The divergence of the split, To match the total number of historical tasks with the same classification identifier, The weighting coefficients for the pre-defined differences in difficulty. For experts For the first The specific score assigned to each historical task. This is the median of all specific scores.

[0012] In a preferred embodiment, the completion events of each branch node in the monitoring scoring task distribution workflow instance include: The workflow engine registers a completion event listener for each branch node in the workflow instance; The node identifier of the branch node and the index of the distribution record of all scoring tasks under the branch node are used as the node completion signal; After completing all the scoring tasks in its input queue, each branch node actively sends a node completion signal to the completion event listener. After the completion event listener receives the node completion signal, it marks the completion status of the corresponding branch node as true and temporarily stores the dispatch record index in the temporary cache of the aggregation node; When the completion status of all branch nodes is true, the completion event listener triggers a summary notification to the aggregation node.

[0013] In a preferred embodiment, the step of summarizing and generating a complete scoring report and pushing the scoring report to the front-end interface of the ETP platform includes: After receiving the summary notification, the aggregation node sends a scoring result extraction request to each branch node; For a branch node that does not return a scoring result within the preset timeout period, the aggregation node queries the historical distribution records of the branch node to obtain the expert identifier and task distribution time assigned to the branch node. The aggregation node sends a status query request for the branch node to the node management interface of the workflow engine. If the query result indicates that the branch node has failed or the task backlog has exceeded the limit, the aggregation node will reassign the scoring task corresponding to the branch node to the backup branch node. Once all branch nodes have successfully returned their scoring results, the aggregation node merges all the returned scoring results according to the task package identifier and evaluation dimension, generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

[0014] To address the aforementioned issues, this invention also provides a one-click scoring task distribution system for an ETP platform based on a workflow engine, the system comprising: The task flow topology scheduling module is used to allocate workflow instances to the scoring task set in the ETP platform according to the workflow engine, activate each branch node in sequence according to the topology order, and pass the scoring task group corresponding to each branch node into the input queue of the currently activated branch node. The intelligent expert matching and distribution module is used to call the expert resource library of the ETP platform for the currently active branch node, dynamically match candidate experts for each scoring task in the input queue of the branch node, and write the distribution record into the distributed log storage of the ETP platform. The results aggregation and report push module is used to listen for the completion events of each branch node in the scoring task distribution workflow instance. When all branch nodes report task completion, the workflow engine's aggregation node collects the scoring results returned by each branch node, summarizes and generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention relies on a workflow engine to realize the instantiation and orchestration of scoring task sets and the topology-driven execution. It can autonomously divide serial and parallel branch nodes according to task dependencies, dynamically construct independent sub-workflow execution contexts and configure appropriate input queue capacity, realize the standardized and automated flow scheduling of scoring tasks, effectively improve the regularity of task distribution, the orderliness of execution and the overall scheduling efficiency, and at the same time ensure the stability and controllability of the parallel processing of multi-branch tasks.

[0016] 2. This invention achieves intelligent and accurate matching of scoring tasks and experts through a multi-dimensional hierarchical screening mechanism. It combines professional expertise, scoring accuracy, task load, and historical scoring characteristics to achieve layer-by-layer optimization. At the same time, it relies on distributed log storage to achieve standardized retention and indexing of distribution records. With the help of full-node event monitoring, result aggregation, and abnormal node fallback redistribution mechanism, it can ensure the rationality of expert matching and scoring consistency, strengthen the traceability of the entire task process, improve the automation level of scoring result aggregation and integration, the timeliness of report generation, and the reliability of the platform's overall scoring business operation. Attached Figure Description

[0017] Figure 1 A flowchart illustrating a one-click scoring task distribution method for an ETP platform based on a workflow engine, provided as an embodiment of the present invention; Figure 2 A functional module diagram of a one-click scoring task distribution system for an ETP platform based on a workflow engine, provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] This application provides a one-click scoring task distribution method for an ETP platform based on a workflow engine. The execution entity of this one-click scoring task distribution method for an ETP platform based on a workflow engine includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application embodiment: a server, a terminal, etc. In other words, this one-click scoring task distribution method for an ETP platform based on a workflow engine can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a one-click scoring task distribution method for an ETP platform based on a workflow engine, according to an embodiment of the present invention. In this embodiment, the one-click scoring task distribution method for an ETP platform based on a workflow engine includes: S1. Based on the workflow engine, assign workflow instances to the scoring task set in the ETP platform, activate each branch node in sequence according to the topology, and pass the scoring task group corresponding to each branch node into the input queue of the currently activated branch node. In this embodiment of the invention, the step of allocating workflow instances to the scoring task set in the ETP platform according to the workflow engine includes: The workflow engine parses the dependency tags between various scoring tasks in the scoring task set of the ETP platform. If there are multiple scoring tasks with dependencies, the workflow engine will orchestrate them as serial branch nodes. If there is no dependency, the workflow engine will divide the scoring task into several parallel task groups according to the preset concurrency threshold, and each parallel task group corresponds to a parallel branch node. The workflow engine generates an independent sub-workflow execution context for each parallel branch node, and records the task identifier list of the scoring tasks contained in the parallel branch node in the sub-workflow execution context; The workflow engine connects serial and parallel branch nodes in topological order to generate a complete workflow definition file, and instantiates the workflow definition file to obtain a scoring task distribution workflow instance.

[0021] The step of activating each branch node sequentially according to the topological order and passing the scoring task group corresponding to each branch node into the input queue of the currently activated branch node includes: The workflow engine reads the node identifier of the starting branch node from the workflow instance and marks the starting branch node as active. For the currently active branch node, the workflow engine dynamically allocates the capacity of the input queue based on the number of tasks in the scoring task group corresponding to the branch node, and pushes each scoring task in the scoring task group into the input queue in turn. Once the currently active branch node has finished processing all scoring tasks in the input queue, the workflow engine queries the topological order of the workflow instance, determines the node identifier of the next branch node, and activates the next branch node, until all branch nodes are activated and processed.

[0022] The workflow engine retrieves the complete scoring task set stored in the ETP platform, breaks down each scoring task in the task set one by one, accurately reads the identification information attached to each task, extracts the dependency relationship tags used to define the sequential relationship between tasks, and establishes a unique binding relationship between each scoring task and the corresponding dependency relationship tag.

[0023] The workflow engine compares and verifies all dependency tags one by one, accurately identifies multiple scoring tasks with sequential constraints, categorizes these scoring tasks with a fixed execution order, and integrates and arranges them in an orderly manner according to the inherent sequential execution logic of the tasks, ultimately solidifying them into serial branch nodes with a complete structure and a clear execution order.

[0024] The workflow engine uniformly aggregates independent scoring tasks that do not have identified dependency tags. According to the system's pre-set concurrency control standards, the aggregated independent scoring tasks are equally divided into several parallel task groups of the same size. Each parallel task group directly corresponds to an independent parallel branch node.

[0025] The workflow engine allocates a dedicated running space for each completed parallel branch node, building a sub-workflow execution context that does not interfere with each other and runs independently. Simultaneously, it enters the unique identity of all scoring tasks contained in the corresponding parallel branch node one by one, completing the complete registration and retention of the task identity list.

[0026] The workflow engine sequentially connects the established serial branch nodes and all parallel branch nodes according to the preset process order logic, splicing them together to form a complete and well-structured workflow definition file. Then, it loads and activates the workflow definition file to complete the instantiation operation, and finally obtains a usable scoring task distribution workflow instance.

[0027] The workflow engine interfaces with the internal information storage area of ​​the scoring task distribution workflow instance, traverses the instance's built-in node information storage blocks layer by layer, locates the first node in the entire process chain, accurately extracts the node identifier of the starting branch node unique to that node, and completely preserves the original ownership and identity information of the identifier. After the identifier is fully read, the workflow instance's built-in node status writing mechanism is invoked to reset the status of the located starting branch node, directly changing the node's initial idle state to an active state that can participate in process scheduling and undertake scoring tasks. At the same time, the updated node status is synchronously and permanently stored in the workflow instance's node status record area.

[0028] The workflow engine matches the scoring task groups bound to the currently active branch nodes, counts and counts all scoring tasks in each task group, and determines the actual total number of scoring tasks. Based on the actual scale of the tasks, it plans and matches dedicated cache space for the branch nodes, and completes the dynamic division and capacity allocation of the input queue storage space. According to the predetermined order within the scoring task group, it retrieves the independent scoring tasks in the group one by one and pushes each scoring task into the configured input queue in a progressive manner, strictly following the original order to complete the enqueueing and storage operation of all scoring tasks.

[0029] The currently active branch node sequentially performs standardized business processing on all scoring tasks stored in the input queue. When there are no remaining scoring tasks to be processed in the input queue, the branch node sends a feedback signal indicating that the task processing is complete to the workflow engine. After receiving the signal, the workflow engine retrieves the pre-organized and fixed topology sequence link data within the workflow instance, searches and locks the adjacent subsequent nodes along the topology flow path, reads the unique node identifier corresponding to the node, and sets the subsequent branch node to the active state according to the unified state configuration rules. The complete process of node retrieval, identifier reading, and state activation is repeated, traversing all branch nodes in the topology link one by one, until each branch node has completed the activation action and all its associated scoring tasks have been processed.

[0030] The beneficial effects include parsing the dependency tags of scoring tasks through the workflow engine, constructing serial and parallel branch nodes as needed, configuring independent sub-workflow execution contexts for parallel branch nodes and recording task identifier lists, and then integrating them according to the topology order to generate workflow definition files and form scoring task distribution workflow instances, thus achieving standardized and structured orchestration of scoring tasks. Simultaneously, it automatically identifies and activates the starting branch node based on the topology order, dynamically matches the input queue capacity according to the size of the scoring task group, and orderly sends scoring tasks into the corresponding input queue. After a single node task is processed, it automatically retrieves the topology order to activate the next node until all nodes are processed, achieving automated operation of scoring task orchestration, instance generation, node activation, task queuing, and sequential flow, improving the orderliness, coherence, and overall operational efficiency of task distribution and scheduling.

[0031] S2. For the currently active branch node, call the expert resource library of the ETP platform to dynamically match candidate experts for each scoring task in the input queue of the branch node, and write the distribution record into the distributed log storage of the ETP platform. In this embodiment of the invention, the step of calling the expert resource library of the ETP platform for the currently active branch node to dynamically match candidate experts for each scoring task in the branch node's input queue includes: Analyze the category identifiers and task difficulty coefficients contained in the current scoring task; Based on the classification identifier, extract the first set of experts with corresponding expertise from the expert resource database; Based on the task difficulty coefficient, experts whose historical scoring accuracy is lower than the preset accuracy threshold are removed from the first expert set to obtain the second expert set; Obtain the number of tasks currently pending for each expert in the second expert set, and sort the second expert set in ascending order of the number of tasks currently pending. Select the expert at the top of the sorted set as the candidate expert for the current scoring task.

[0032] The process of dynamically matching candidate experts for each scoring task in the branch node input queue also includes: After selecting candidate experts, a second screening is conducted based on the experts' historical scoring preferences. Extract the distribution characteristics of the historical scores given by each candidate expert for past tasks with the same classification label as the current scoring task from the historical scoring records of the ETP platform. The scoring divergence of each candidate expert is calculated based on the distribution characteristics, and the scoring divergence is compared with the scoring consistency requirements of the current scoring task. Candidate experts whose scoring dispersion exceeds the scoring consistency requirement are removed. The remaining candidate experts are then sorted according to their historical collaboration frequency with the task initiator of the current scoring task, and the expert ranked first is selected as the final matching expert.

[0033] The step of writing the distribution records into the distributed log storage of the ETP platform includes: The process record for matching candidate experts is integrated with the unique identifier of the scoring task, the identifier of the matched candidate expert, the distribution timestamp, and the node identifier of the branch node into a distribution record; The distribution record is encapsulated into a log entry, and the write interface of the distributed log storage is called to write the log entry to the currently active shard of the distributed log storage. After a log entry is successfully written, the distributed log storage returns a write confirmation message to the workflow engine and synchronizes the storage location index of the log entry to the audit log index table of the ETP platform.

[0034] The splitting divergence includes: ; In the formula, For experts The divergence of the split, To match the total number of historical tasks with the same classification identifier, The weighting coefficients for the pre-defined differences in difficulty. For experts For the first The specific score assigned to each historical task. This is the median of all specific scores.

[0035] The workflow engine performs a structured decomposition and reading of each scoring task in the input queue of the branch node, parses the core identifier fields built into the task line by line, accurately extracts the classification identifier and task difficulty coefficient used to match experts, and binds and stores the two types of identifier information with the unique identity information of the current scoring task to ensure that the correspondence between the identifier information and the task is not deviated, providing a clear basis for subsequent expert matching.

[0036] Using the parsed classification identifier as the sole retrieval criterion, the system connects to the expert resource library built into the ETP platform, traverses all expert specialty registration files stored in the resource library one by one, filters out all experts whose specialty fields are completely consistent with the classification identifier of the current scoring task, and unifies and integrates all the filtered experts to form a complete first expert set.

[0037] Based on the analyzed task difficulty coefficient, retrieve the historical scoring records archived by each expert in the first expert set, verify the degree of fit between each expert's past scoring results and the standard results, determine whether their historical scoring accuracy meets the established qualification standard, completely remove experts who do not meet the standard from the first expert set, and re-collect and organize the remaining qualified experts to form the second expert set.

[0038] The system retrieves the pending task registration information of each expert in the second expert set, which is synchronized to the platform in real time. It then counts and records the number of pending tasks for each expert, compares the number of pending tasks for all experts horizontally, and ranks the second expert set in a fixed order from fewest to most tasks. The expert ranked first in the ranking is directly identified as the candidate expert for the current scoring task.

[0039] The workflow engine identifies the candidate expert group after the initial matching and, relying on the expert behavior archiving mechanism built into the ETP platform, retrieves the long-term scoring behavior records of each candidate expert. Based on the fixed behavioral tendencies in the records, it completes the preference classification in the professional dimension. The classification results are used as the screening benchmark to carry out a second round of targeted screening of the candidate expert group, further narrowing the scope of experts who can participate in the current scoring task from the perspective of professional suitability.

[0040] The workflow engine accesses the ETP platform's dedicated historical scoring record storage area, locates all historical task entries with the same category identifier as the current scoring task, retrieves all scoring content issued by the corresponding candidate experts for the same type of historical task, and comprehensively sorts and summarizes the batch scoring content to extract content that can objectively reflect the scoring pattern, such as the overall clustering status of scores and the range of score fluctuations, thus forming a complete distribution characteristic of the historical scores for each candidate expert.

[0041] The workflow engine relies on the distribution characteristics of the extracted historical scores to analyze the fluctuations in the scoring content of each candidate expert in previous rounds. This quantitatively determines the stability of the expert's scoring behavior and forms the scoring divergence. It then retrieves the pre-set scoring consistency requirements standard for the current scoring task and compares the scoring divergence of each candidate expert with the established standard one by one to clarify the compatibility between the two.

[0042] The workflow engine compares the results and eliminates candidate experts whose dispersion exceeds the scoring consistency requirements. It retains qualified candidate experts who meet the consistency requirements, retrieves the task collaboration archive data stored in the platform, counts the number of times each qualified candidate expert has cooperated with the task initiator of the current scoring task, and ranks them according to the frequency of cooperation. The expert ranked first in the ranking sequence is selected as the final matching expert for the current scoring task.

[0043] The workflow engine fully collects the process records generated throughout the entire expert matching process. At the same time, it retrieves the unique identifier of the current scoring task, the identifier of the selected candidate expert after the matching is completed, the distribution timestamp of the task matching completion time, and the node identifier of the current branch node. All the collected and retrieved information is merged and collected according to a fixed structure specification, arranged in a standardized manner according to the preset information combination format, and uniformly integrated and assembled into a standardized structured distribution record.

[0044] The workflow engine standardizes and encapsulates the integrated distribution records, formats and adapts the content according to the item structure specifications of the distributed log storage, generates log items that conform to the storage specifications, actively connects to the pre-set communication interaction channel of the distributed log storage, calls the dedicated write interface configured by the distributed log storage, locates the storage shards of the distributed log storage that are in a working state in real time, and transmits and writes the encapsulated log items to the currently active shard of the distributed log storage to complete the final storage.

[0045] After the distributed log storage completes the writing and storage of log entries, it automatically generates a feedback message that identifies the writing result, pushes a fixed-format write confirmation message to the workflow engine, synchronously extracts the storage location information of log entries in the current active shard, organizes and generates a dedicated storage location index, and enters the storage location index one by one into the pre-set audit log index table of the ETP platform to complete the registration and filing, so as to facilitate the subsequent retrieval of the corresponding log entries and distribution records at any time.

[0046] The total number of historical tasks matching the same classification identifier is derived from experts on the ETP platform. All historical scoring task records that match the current task category identifier. The preset weighting coefficients for difficulty differences are derived from a fixed weighting configuration table pre-configured by the ETP platform for historical tasks of different difficulty levels. Expert The specific score assigned to each similar historical task comes from the expert data stored in the historical scoring records of the ETP platform. The actual scores for each similar historical task. The median of all specific scores is derived from expert evaluation. The result is obtained by sorting the specific scores of all similar historical tasks and taking the middle value.

[0047] Extraction Expert For all historical tasks with the same category identifier as the current scoring task, calculate the difference between each specific score and the median of all specific scores. Then, square the difference. Multiply the result of each square by a preset weighting coefficient for the corresponding difficulty difference. Sum all the multiplicative results and sum them together. Divide the sums by the summation and take the square root of the result. The result is the expert score. The divergence of the split.

[0048] Divergence is used to measure the expertise of experts The value indicates the degree of fluctuation in the scoring results for tasks with the same classification label. The larger the value, the more obvious the fluctuation in the expert scoring results. The smaller the value, the more stable the expert scoring results. It can be directly used to determine whether the expert scoring behavior meets the scoring consistency requirements of the current scoring task.

[0049] The beneficial effects include achieving dynamic and precise matching of scoring tasks and experts through a multi-dimensional, hierarchical screening mechanism. Targeted optimization is achieved based on classification labels, historical scoring performance, task load, scoring dispersion, and collaboration frequency, ensuring expert expertise matching, stable scoring, and balanced load, effectively improving the professionalism and consistency of scoring results. Simultaneously, process records and key identifiers are integrated to generate distribution records, which are encapsulated and written into distributed log storage with synchronized audit indexes. This enables standardized log retention and traceable management of the entire task distribution process, ensuring complete and verifiable task flow information. This provides a reliable basis for subsequent auditing, tracing, and anomaly investigation, improving the standardization and controllability of platform operations.

[0050] S3. Listen for the completion events of each branch node in the scoring task distribution workflow instance. When all branch nodes report task completion, the workflow engine's aggregation node collects the scoring results returned by each branch node, summarizes and generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

[0051] In this embodiment of the invention, the completion events of each branch node in the monitoring scoring task distribution workflow instance include: The workflow engine registers a completion event listener for each branch node in the workflow instance; The node identifier of the branch node and the index of the distribution record of all scoring tasks under the branch node are used as the node completion signal; After completing all the scoring tasks in its input queue, each branch node actively sends a node completion signal to the completion event listener. After the completion event listener receives the node completion signal, it marks the completion status of the corresponding branch node as true and temporarily stores the dispatch record index in the temporary cache of the aggregation node; When the completion status of all branch nodes is true, the completion event listener triggers a summary notification to the aggregation node.

[0052] The process of generating a complete scoring report and pushing it to the front-end interface of the ETP platform includes: After receiving the summary notification, the aggregation node sends a scoring result extraction request to each branch node; For a branch node that does not return a scoring result within the preset timeout period, the aggregation node queries the historical distribution records of the branch node to obtain the expert identifier and task distribution time assigned to the branch node. The aggregation node sends a status query request for the branch node to the node management interface of the workflow engine. If the query result indicates that the branch node has failed or the task backlog has exceeded the limit, the aggregation node will reassign the scoring task corresponding to the branch node to the backup branch node. Once all branch nodes have successfully returned their scoring results, the aggregation node merges all the returned scoring results according to the task package identifier and evaluation dimension, generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

[0053] The workflow engine performs a full traversal and analysis of the currently running workflow instances, identifies each branch node configured and deployed within the workflow instance, configures a dedicated listener binding relationship for each identified branch node, and registers a dedicated and independent completion event listener for each branch node, so that each branch node has a dedicated matching event listener receiving subject.

[0054] Collect the unique node identifier information of each branch node, then sort out all the scoring tasks under the jurisdiction of the branch node one by one, extract the distribution record index corresponding to each scoring task, integrate and encapsulate the node identifier of the branch node and the distribution record index corresponding to all scoring tasks, and form a unique node completion signal after fixing the combination format.

[0055] The branch node continuously reads the pending content in its configured input queue and executes each scoring task in the input queue in sequence until all scoring tasks in the input queue have been completed and there is no remaining pending content. The branch node actively establishes a communication channel with its bound completion event listener, and sends the node completion signal directly after properly encapsulating the preset format.

[0056] The event listener keeps a signal receiving standby state at all times, accurately receives the node completion signal sent by the corresponding branch node, disassembles and analyzes the information content carried inside the signal, matches and locates the corresponding branch node, directly marks the completion status of the branch node as true, and extracts the distribution record index contained in the signal, and completely transfers it to the temporary cache space pre-divided by the aggregation node for temporary storage.

[0057] The completion event listener continuously polls and verifies the status flag information of all branch nodes under the workflow instance, checking the completion status value of each branch node one by one. When the completion status of all branch nodes is consistent with true value, the completion event listener immediately initiates an event triggering action to the aggregation node and directly pushes the corresponding summary notification to the aggregation node.

[0058] The aggregation node listens to the summary notifications pushed by the system in real time. After fully receiving and parsing the entire content of the summary notification, it sorts out all the branch nodes currently involved in the task processing one by one, establishes communication connections in sequence according to the unique identifier of the node, and sends the corresponding scoring result extraction request to each branch node precisely to ensure that each branch node can receive the exclusive request instruction.

[0059] After the aggregation node sends a scoring result retrieval request to each branch node, it continuously monitors the result return response of each branch node. If a branch node fails to return the corresponding scoring result within the preset timeout period, the aggregation node immediately retrieves the historical distribution record of that branch node stored in the system backend, accurately retrieves and obtains the expert identification information that has been pre-assigned and bound to that branch node from the record, and simultaneously extracts the accurate task distribution time corresponding to this task.

[0060] The aggregation node organizes the identity code information of the abnormal and unresponsive branch nodes, connects to the node management interface configured by the workflow engine according to the established communication specifications, initiates a dedicated status query request for the abnormal branch node to the interface, receives the node running status data fed back by the interface, and strictly determines whether the branch node has experienced a running failure or whether the internal task backlog has reached the limit standard based on the feedback content. After the determination is confirmed, all the scoring tasks undertaken by the branch node are directly redistributed and scheduled to the pre-configured backup branch nodes for processing.

[0061] After all branch nodes successfully transmit their processed scoring results back to the aggregation node as required, the aggregation node reads each scoring result one by one, classifies and collects them according to the task package identifier, and then organizes and merges the collected scoring results in an orderly manner according to the preset evaluation dimensions. After organizing all the sub-evaluation data, a complete scoring report with standardized format and complete content is generated. The aggregation node establishes a dedicated data push link and directly pushes the generated complete scoring report to the corresponding front-end interface of the ETP platform for page display.

[0062] The beneficial effects include: registering dedicated completion event listeners for each branch node, and combining this with node completion signals to achieve precise real-time monitoring of the task status of branch nodes. This automatically identifies task completion progress and triggers summary notifications without manual intervention. Upon receiving the notification, the aggregation node extracts the scoring results, automatically traces historical records and verifies the running status of nodes that have timed out, and reassigns the scoring tasks of abnormal nodes to backup nodes, avoiding task delays and data loss. Finally, the results are merged according to task package identifiers and evaluation dimensions to generate a complete scoring report, which is then pushed to the ETP platform frontend. This achieves full-process automation, improves task processing efficiency and stability, and ensures complete and timely display of scoring data.

[0063] like Figure 2 The diagram shown is a functional block diagram of a one-click scoring task distribution system for an ETP platform based on a workflow engine, provided by an embodiment of the present invention.

[0064] The workflow engine-based ETP platform one-click scoring task distribution system 100 described in this invention can be installed in an electronic device. Depending on the functions implemented, the workflow engine-based ETP platform one-click scoring task distribution system 100 may include a task flow topology scheduling module 101, an intelligent expert matching and distribution module 102, and a result aggregation and report push module 103. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0065] In this embodiment, the functions of each module / unit are as follows: The task flow topology scheduling module 101 is used to allocate workflow instances to the scoring task set in the ETP platform according to the workflow engine, activate each branch node in sequence according to the topology order, and pass the scoring task group corresponding to each branch node into the input queue of the currently activated branch node. The intelligent expert matching and distribution module 102 is used to call the expert resource library of the ETP platform for the currently active branch node, dynamically match candidate experts for each scoring task in the input queue of the branch node, and write the distribution record into the distributed log storage of the ETP platform. The result aggregation and report push module 103 is used to listen to the completion events of each branch node in the scoring task distribution workflow instance. When all branch nodes report task completion, the aggregation node of the workflow engine collects the scoring results returned by each branch node, summarizes and generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

[0066] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0067] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0068] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0069] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0070] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A one-click scoring task distribution method for an ETP platform based on a workflow engine, characterized in that, The method includes: S1. Based on the workflow engine, assign workflow instances to the scoring task set in the ETP platform, activate each branch node in sequence according to the topology, and pass the scoring task group corresponding to each branch node into the input queue of the currently activated branch node. S2. For the currently active branch node, call the expert resource library of the ETP platform to dynamically match candidate experts for each scoring task in the input queue of the branch node, and write the distribution record into the distributed log storage of the ETP platform. S3. Listen for the completion events of each branch node in the scoring task distribution workflow instance. When all branch nodes report task completion, the workflow engine's aggregation node collects the scoring results returned by each branch node, summarizes and generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

2. The one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 1, characterized in that, The process of allocating workflow instances to the scoring task set in the ETP platform according to the workflow engine includes: The workflow engine parses the dependency tags between various scoring tasks in the scoring task set of the ETP platform. If there are multiple scoring tasks with dependencies, the workflow engine will orchestrate them as serial branch nodes. If there is no dependency, the workflow engine will divide the scoring task into several parallel task groups according to the preset concurrency threshold, and each parallel task group corresponds to a parallel branch node. The workflow engine generates an independent sub-workflow execution context for each parallel branch node, and records the task identifier list of the scoring tasks contained in the parallel branch node in the sub-workflow execution context; The workflow engine connects serial and parallel branch nodes in topological order to generate a complete workflow definition file, and instantiates the workflow definition file to obtain a scoring task distribution workflow instance.

3. The one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 1, characterized in that, The step of activating each branch node sequentially according to the topological order and passing the scoring task group corresponding to each branch node into the input queue of the currently activated branch node includes: The workflow engine reads the node identifier of the starting branch node from the workflow instance and marks the starting branch node as active. For the currently active branch node, the workflow engine dynamically allocates the capacity of the input queue based on the number of tasks in the scoring task group corresponding to the branch node, and pushes each scoring task in the scoring task group into the input queue in turn. Once the currently active branch node has finished processing all scoring tasks in the input queue, the workflow engine queries the topological order of the workflow instance, determines the node identifier of the next branch node, and activates the next branch node, until all branch nodes are activated and processed.

4. The one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 1, characterized in that, For the currently active branch node, the ETP platform's expert resource library is invoked to dynamically match candidate experts for each scoring task in the branch node's input queue, including: Analyze the category identifiers and task difficulty coefficients contained in the current scoring task; Based on the classification identifier, extract the first set of experts with corresponding expertise from the expert resource database; Based on the task difficulty coefficient, experts whose historical scoring accuracy is lower than the preset accuracy threshold are removed from the first expert set to obtain the second expert set; Obtain the number of tasks currently pending for each expert in the second expert set, and sort the second expert set in ascending order of the number of tasks currently pending. Select the expert at the top of the sorted set as the candidate expert for the current scoring task.

5. The one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 4, characterized in that, The process of dynamically matching candidate experts for each scoring task in the branch node input queue also includes: After selecting candidate experts, a second screening is conducted based on the experts' historical scoring preferences. Extract the distribution characteristics of the historical scores given by each candidate expert for past tasks with the same classification label as the current scoring task from the historical scoring records of the ETP platform. The scoring divergence of each candidate expert is calculated based on the distribution characteristics, and the scoring divergence is compared with the scoring consistency requirements of the current scoring task. Candidate experts whose scoring dispersion exceeds the scoring consistency requirement are removed. The remaining candidate experts are then sorted according to their historical collaboration frequency with the task initiator of the current scoring task, and the expert ranked first is selected as the final matching expert.

6. The one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 5, characterized in that, The step of writing the distribution records into the distributed log storage of the ETP platform includes: The process record for matching candidate experts is integrated with the unique identifier of the scoring task, the identifier of the matched candidate expert, the distribution timestamp, and the node identifier of the branch node into a distribution record; The distribution record is encapsulated into a log entry, and the write interface of the distributed log storage is called to write the log entry to the currently active shard of the distributed log storage. After a log entry is successfully written, the distributed log storage returns a write confirmation message to the workflow engine and synchronizes the storage location index of the log entry to the audit log index table of the ETP platform.

7. The one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 5, characterized in that, The splitting divergence includes: ； In the formula, For experts The divergence of the split, To match the total number of historical tasks with the same classification identifier, The weighting coefficients for the pre-defined differences in difficulty. For experts For the first The specific score assigned to each historical task. This is the median of all specific scores.

8. The one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 1, characterized in that, The completion events of each branch node in the monitoring and scoring task distribution workflow instance include: The workflow engine registers a completion event listener for each branch node in the workflow instance; The node identifier of the branch node and the index of the distribution record of all scoring tasks under the branch node are used as the node completion signal; After completing all the scoring tasks in its input queue, each branch node actively sends a node completion signal to the completion event listener. After the completion event listener receives the node completion signal, it marks the completion status of the corresponding branch node as true and temporarily stores the dispatch record index in the temporary cache of the aggregation node; When the completion status of all branch nodes is true, the completion event listener triggers a summary notification to the aggregation node.

9. The one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 8, characterized in that, The process of generating a complete scoring report and pushing it to the front-end interface of the ETP platform includes: After receiving the summary notification, the aggregation node sends a scoring result extraction request to each branch node; For a branch node that does not return a scoring result within the preset timeout period, the aggregation node queries the historical distribution records of the branch node to obtain the expert identifier and task distribution time assigned to the branch node. The aggregation node sends a status query request for the branch node to the node management interface of the workflow engine. If the query result indicates that the branch node has failed or the task backlog has exceeded the limit, the aggregation node will reassign the scoring task corresponding to the branch node to the backup branch node. Once all branch nodes have successfully returned their scoring results, the aggregation node merges all the returned scoring results according to the task package identifier and evaluation dimension, generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

10. A one-click scoring task distribution system for an ETP platform based on a workflow engine, characterized in that, The system is used to implement the one-click scoring task distribution method for an ETP platform based on a workflow engine as described in claim 1, the system comprising: The task flow topology scheduling module is used to allocate workflow instances to the scoring task set in the ETP platform according to the workflow engine, activate each branch node in sequence according to the topology order, and pass the scoring task group corresponding to each branch node into the input queue of the currently activated branch node. The intelligent expert matching and distribution module is used to call the expert resource library of the ETP platform for the currently active branch node, dynamically match candidate experts for each scoring task in the input queue of the branch node, and write the distribution record into the distributed log storage of the ETP platform. The results aggregation and report push module is used to listen for the completion events of each branch node in the scoring task distribution workflow instance. When all branch nodes report task completion, the workflow engine's aggregation node collects the scoring results returned by each branch node, summarizes and generates a complete scoring report, and pushes the scoring report to the front-end interface of the ETP platform.

Images