Task chain construction method and system for skill training platform, and medium

Abstract

The invention provides a task chain construction method and system oriented to a skill training platform and a medium, and the method comprises the steps: obtaining an external demand data stream associated with a target object identifier in the skill training platform, the external demand data stream comprising a plurality of skill demand description texts with time sequence marks; performing entity extraction processing on the external demand data stream to obtain a plurality of skill demand feature entities; determining a demand weight value of each skill demand feature entity according to the occurrence frequency and the time sequence mark of the skill demand feature entity in the external demand data stream; taking the skill demand feature entities as atlas nodes, and writing the demand weight values into attribute parameters of the corresponding atlas nodes to obtain a skill demand feature atlas; and matching a corresponding target task unit based on the skill demand feature map, and further generating a target task chain for the target object identifier. According to the method, the constructed task chain and the real-time changing external skill demand environment can be kept synchronous, and the timeliness of the task chain is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of skill training, and particularly relates to a task chain construction method and system for a skill training platform and a medium. BACKGROUND

[0002] In the skill training platform, constructing a task chain based on target skill requirements is a key technology for realizing automatic resource management. The construction of the task chain usually relies on a pre-set relational database, which stores fixed mapping relationships between skill characteristic attributes and task units. For example, by querying the pre-set fixed mapping relationships, the corresponding task units can be matched, and then the task chain can be constructed.

[0003] However, the timeliness of the task chain constructed based on the fixed mapping relationships is usually poor, which cannot meet the current actual skill requirements. SUMMARY

[0004] Embodiments of the present application provide a task chain construction method, system and medium for a skill training platform, aiming to improve the timeliness of the constructed task chain.

[0005] In a first aspect, embodiments of the present application provide a task chain construction method for a skill training platform, which comprises:

[0006] Obtaining external demand data flow associated with a target object identifier in a skill training platform, wherein the external demand data flow includes a plurality of skill requirement description texts with time sequence marks;

[0007] Performing entity extraction processing on the external demand data flow to obtain a plurality of skill requirement characteristic entities;

[0008] Determining a requirement weight value of each skill requirement characteristic entity according to the occurrence frequency of the skill requirement characteristic entity in the external demand data flow and the time sequence mark;

[0009] Taking the skill requirement characteristic entity as a graph node, and writing the requirement weight value into the attribute parameter of the corresponding graph node to obtain a skill requirement characteristic graph;

[0010] Based on the skill requirement characteristic graph, matching a corresponding target task unit from a pre-set task unit library;

[0011] Generating a target task chain for the target object identifier according to the target task unit.

[0012] In the above embodiment, by collecting and analyzing external timing data in real time, a dynamically weighted skill demand feature map is constructed, and abstract market skill demand is automatically mapped to specific execution task units. This scheme can eliminate information asymmetry, ensure that the generated execution sequence is synchronized with the real-time changing external skill demand environment, realize the automatic and precise derivation from macro target to micro operation path, and effectively improve the scientificity and immediate adaptability of the execution plan.

[0013] In an embodiment, the determination of the demand weight value of each skill demand feature entity according to the occurrence frequency of the skill demand feature entity in the external demand data stream and the timing mark comprises:

[0014] determining a time difference value between the timing mark and the current system time;

[0015] determining a time decay coefficient of the corresponding skill demand feature entity according to the time difference value;

[0016] determining the demand weight value of the corresponding skill demand feature entity based on the time decay coefficient and the occurrence frequency.

[0017] In the above embodiment, by introducing dynamic decay calculation logic based on time difference value, the frequency data is weighted and corrected by using time decay coefficient, so that the generated feature map can accurately reflect the real skill demand heat of the current time section, effectively eliminating the interference of outdated information, and ensuring that the subsequent constructed task chain has high market sensitivity and timeliness value.

[0018] In an embodiment, the matching of the corresponding target task unit from the preset task unit library based on the skill demand feature map comprises:

[0019] determining a target graph node in the skill demand feature map;

[0020] determining, for each preset task unit in the task unit library, a content coverage index, a difficulty gradient matching degree, and a historical contribution effectiveness coefficient between the preset task unit and the target graph node;

[0021] determining a matching score between the preset task unit and the target graph node based on the content coverage index, the difficulty gradient matching degree, and the historical contribution effectiveness coefficient;

[0022] determining the target task unit from the task unit library according to the matching score.

[0023] In the above embodiment, by constructing a multi-dimensional evaluation system including semantic coverage, difficulty adaptation and historical performance, accurate mapping between task units and demand nodes is realized. This scheme not only ensures the consistency of the selected tasks in knowledge points, but also ensures the executability and effectiveness of the tasks through the introduction of difficulty matching and historical data verification, thereby generating high-quality and high-conversion execution sequences.

[0024] In an embodiment, the target task unit is determined from the task unit library according to the matching score, including:

[0025] Based on the demand weight value of the target graph node, the target task quantity is determined, and the target task quantity is positively correlated with the demand weight value of the target graph node;

[0026] According to the matching score, the target task unit matched with the target task quantity is determined from the task unit library.

[0027] In the above embodiment, by establishing a positive correlation mapping mechanism between task quantity and feature weight, automation and intelligentization of resource allocation are realized. This scheme can dynamically adjust the execution load according to the urgency of market skill demand, automatically increase the training proportion for high-value core skills, and simplify the edge skills, thereby ensuring the construction of core competitiveness while effectively optimizing the overall execution efficiency and time cost.

[0028] In an embodiment, the target task chain for the target object identifier is generated according to the target task unit, including:

[0029] A directed acyclic graph including the target task unit is constructed;

[0030] The historical execution log data of the target object identifier is obtained, and the historical completed tasks are included in the historical execution log data;

[0031] Based on the historical completed tasks, the directed acyclic graph is pruned to obtain a processed directed acyclic graph;

[0032] According to the processed directed acyclic graph, the target task chain for the target object identifier is generated.

[0033] In the above embodiment, by constructing a task dependency graph and dynamically pruning combined with user historical data, intelligent deduplication and personalized customization of execution path are realized. This scheme can automatically identify and eliminate historical tasks completed by the user, avoid resource waste caused by repeated training, ensure that the generated task chain always focuses on the user's ability gap, and significantly improve the accuracy and execution efficiency of task planning.

[0034] In an embodiment, the generating a target task chain for the target object identifier according to the processed directed acyclic graph comprises:

[0035] Obtaining an estimated execution duration of each node in the processed directed acyclic graph;

[0036] Mapping the estimated execution duration to a weight value of a corresponding node in the directed acyclic graph;

[0037] Using Dijkstra algorithm to determine a target path sequence from a start node to an end node in the directed acyclic graph, with an optimization objective of minimizing the sum of weight values on the path;

[0038] Generating a target task chain for the target object identifier based on the target path sequence.

[0039] In the above embodiment, by introducing time dimension as a core cost function and using shortest path algorithm to automatically search for an optimal solution in a complex task dependency network, a time efficiency oriented task chain construction is realized. The scheme can automatically plan an execution path with the shortest total time consumption and the highest efficiency in the case of multiple technical routes or execution branches, reduces the time cost of the user, and maximizes the execution benefit.

[0040] In an embodiment, after the generating a target task chain for the target object identifier according to the target task unit, the method further comprises:

[0041] Receiving execution feedback data for a current task execution node in the target task chain, the execution feedback data comprising a task execution completion degree score;

[0042] Based on the task execution completion degree score, performing task execution node insertion processing or task execution node deletion processing on the target task chain.

[0043] In the above embodiment, by introducing a dynamic link reconstruction logic based on real-time feedback, adaptive matching of the task chain to the actual ability of the user is realized. The scheme can automatically insert a tutoring task to reduce the gradient when the user performs poorly, and automatically remove redundant tasks to improve efficiency when the user performs well, thereby converting a static plan into a dynamic personalized guidance path, ensuring the smoothness and efficiency of the execution process.

[0044] In an embodiment, the performing task execution node insertion processing or task execution node deletion processing on the target task chain based on the task execution completion degree score comprises at least one of the following steps:

[0045] If the execution completion degree score is less than a preset first threshold value, a to-be-supplemented task unit matched with the current task execution node is obtained from the task unit library, and based on the to-be-supplemented task unit, task execution node insertion processing is performed on the target task chain;

[0046] If the execution completion degree score is greater than a preset second threshold value, a to-be-skipped task execution node located after the current task execution node and having a task difficulty level less than a preset level threshold value is determined in the target task chain, and based on the to-be-skipped task execution node, task execution node deletion processing is performed on the target task chain, and the second threshold value is greater than or equal to the first threshold value.

[0047] In the above embodiment, by automatically inserting a degraded tutorial task when the score is low and automatically removing a low-level redundant task when the score is high, the task chain can be scaled according to the real-time performance of the user. This not only effectively solves the problem of user lag caused by weak foundation, but also avoids the waste of time of high-level users due to repeated low-difficulty exercises, maximizing the ability improvement output per unit of time.

[0048] In a second aspect, the embodiments of the present application provide a task chain construction system for a skill training platform, which is used to execute the task chain construction method for the skill training platform as described in any of the above.

[0049] In a third aspect, the embodiments of the present application provide a computer readable storage medium storing a computer program configured to be executed by a processor to implement the task chain construction method for the skill training platform as described in any of the above.

[0050] The beneficial effects of the embodiments of the present application are as follows:

[0051] In the embodiments of the present application, by performing entity extraction processing on the external demand data stream, a plurality of skill demand feature entities are obtained, the demand weight value of each skill demand feature entity is determined according to the occurrence frequency and time sequence mark of the skill demand feature entity in the external demand data stream, and then a dynamically weighted skill demand feature map is constructed and mapped to a specific target task unit to generate a target task chain, so as to eliminate information asymmetry and ensure that the generated target task chain is synchronized with the real-time changing external skill demand environment, realize accurate derivation from macro external skill demand to micro target task chain, and improve the timeliness and accuracy of the task chain. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0053] Figure 1 is an embodiment flow diagram of a task chain construction method for a skill training platform provided by the embodiments of the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present application. In addition, in the description of the present application, the meaning of "multiple" is two or more, unless otherwise explicitly and specifically limited.

[0055] In a first aspect, the embodiments of the present application provide a task chain construction method for a skill training platform, and the execution subject is a task chain construction system for a skill training platform (hereinafter referred to as "system").

[0056] Specifically, with reference to Figure 1 , the task chain construction method for a skill training platform can include:

[0057] S101, acquiring external demand data stream associated with target object identifier in the skill training platform, the external demand data stream including multiple skill demand description texts with time sequence marks;

[0058] In the embodiments of the present application, the skill training platform refers to a computer software system providing online programming practice, virtual simulation operation, professional skill assessment, and other skill training functions related to professional skill training. The target object identifier refers to a digital fingerprint or character sequence used to uniquely mark the subject of data processing request, which is associated with a specific target field label, such as front-end development engineer or data analyst in computer software engineering, etc. The external demand data stream refers to a collection of unstructured data that is real-time captured or batch imported from third-party data sources (such as recruitment platform open interface, industry technology forum, or project bidding system). The time sequence mark refers to the specific timestamp of each data generation or publication, which is used to indicate the timeliness of the data. The skill demand description text refers to a text passage that describes the job capability, technical stack requirement, or project delivery standard in natural language, such as a character stream containing "mastering Python multithreading programming" or "familiar with Docker containerized deployment".

[0059] In some embodiments of the present application, the acquisition process is performed through a preset application programming interface (API) or web crawler. The system initiates a retrieval request to the distributed network nodes according to the target object identifier mapped to the keyword (such as “Java architect”), filters out duplicate data and invalid data, and retains data within a preset time window (such as 30 days) to form an external demand data stream with high timeliness. Through this step, the system can capture the latest technical trends in the industry, solving the defect that static databases cannot reflect real-time market changes.

[0060] S102, performing entity extraction processing on the external demand data stream to obtain a plurality of skill demand feature entities;

[0061] In embodiments of the present application, entity extraction processing refers to a process of identifying and extracting words or phrases with specific meanings from unstructured text using natural language processing (NLP) technology. Skill demand feature entities refer to the smallest semantic units representing specific technical abilities, tool names, framework names, or theoretical concepts, i.e., “skill atoms”. For example, “Spring Boot” and “microservices” are extracted as skill demand feature entities from “can use Spring Boot to build microservices”.

[0062] In some embodiments of the present application, entity extraction processing can be performed using a deep learning-based named entity recognition (NER) model. The system preloads a technical dictionary for a specific field, which can identify professional terms and their synonym variants (e.g., normalizing “JS” and “JavaScript” to the same entity). Through entity extraction processing, ambiguous text descriptions are converted into computer-calculable structured symbols, providing basic data dimensions for subsequent quantitative analysis.

[0063] S103, determining a demand weight value for each skill demand feature entity according to the occurrence frequency and time sequence label of the skill demand feature entity in the external demand data stream;

[0064] In embodiments of the present application, occurrence frequency refers to the number of times a particular skill demand feature entity is mentioned in the collected sample set, which reflects the market popularity or popularity of the skill. The demand weight value is a quantitative value representing the importance of the entity in achieving the goal.

[0065] S104, taking the skill demand feature entity as a graph node, and writing the demand weight value into an attribute parameter of the corresponding graph node to obtain a skill demand feature graph;

[0066] In an embodiment of the present application, the skill demand feature graph is a data structure for storing knowledge and data relationships in a graph structure, and contains nodes and edges. The graph node is a storage form of the skill demand feature entity in the graph database. The attribute parameter refers to the key-value pair data stored in the node, including but not limited to the weight value, the ability level requirement (such as primary, intermediate, and advanced) and the classification of the technical field.

[0067] In some embodiments of the present application, the construction process of the skill demand feature graph includes creating a virtual root node and connecting all extracted skill demand feature entities as child nodes to the root node, or establishing connection edges between entities according to the co-occurrence relationship between entities. The calculated demand weight value is written into the metadata field of the node, realizing the conversion from unstructured text to a weighted graph. The weighted graph intuitively shows the skill panorama required to achieve the goal and the priority distribution, providing a weighted search space for subsequent path planning.

[0068] S105, based on the skill demand feature graph, matching a corresponding target task unit from a preset task unit library;

[0069] In an embodiment of the present application, the preset task unit library is a database in the skill training platform that stores specific execution instructions, operation items or exercise questions. Each unit has clear input and output standards and assessment points. The target task unit refers to a specific execution action selected to cover a specific skill demand feature entity, such as "write a Hello World program based on Flask".

[0070] In some embodiments of the present application, the matching process of the target task unit is not a simple keyword matching, but a similarity calculation based on semantic vector space or multi-dimensional feature alignment. The system traverses the high-weight nodes in the skill demand feature graph and retrieves tasks in the task unit library that can cover the knowledge points represented by the nodes. For example, for the high-weight graph node "asynchronous programming", the system will match the specific task unit "use Python asyncio library to implement concurrent requests". This step realizes the mapping from "abstract ability requirement" to "specific execution action", solving the problem that the user knows what to learn but does not know how to practice.

[0071] S106, generating a target task chain for the target object identifier according to the target task unit.

[0072] In the embodiments of the present application, the target task chain refers to a sequence of task units arranged in a specific logical order, usually in the form of a linear list or a directed acyclic graph structure, representing the operation path required by the user to reach the target state from the current state.

[0073] In some embodiments of the present application, the generation process of the target task chain includes dependency analysis on the screened multiple target task units. The system reads the preconditions in the task unit metadata, sorts the tasks according to the logical dependency relationship (such as mastering syntax before writing algorithms), and forms an executable sequence. At the same time, according to the weight value of the graph node, the system can arrange the core task with high weight on the core path, ensuring that the user first masters the most urgently needed skills in the market.

[0074] As can be seen, the embodiments of the present application construct a dynamically weighted skill demand feature graph by collecting and analyzing external time series data in real time, and automatically map abstract market skill demand to specific execution task units. This scheme can eliminate information asymmetry, ensure that the generated execution sequence is synchronized with the real-time changing external skill demand environment, realize the automatic and precise derivation from macro target to micro operation path, and effectively improve the scientificity and immediate adaptability of the execution plan.

[0075] In some embodiments of the present application, the demand weight value of each skill demand feature entity is determined according to the occurrence frequency and time sequence label of the skill demand feature entity in the external demand data stream, including:

[0076] S201, determining a time difference value between the time sequence label and the current system time;

[0077] In the embodiments of the present application, the current system time refers to the standard clock time of the server side when performing data processing operations. The time difference value refers to the interval length between the publishing or updating time of each data in the external demand data stream and the current time, usually in days or hours.

[0078] In some embodiments of the present application, the system obtains the time sequence label by analyzing the metadata field of the data packet header, and obtains the difference value by subtraction operation with the system clock. For example, even if a certain technology frequently appears in documents three years ago, if its time difference value is too large, it means that the data belongs to obsolete information. This step provides a basic variable for subsequent weighted calculation with time dimension, ensuring that the system can identify the information value decay caused by time lapse.

[0079] S202, determining a time decay coefficient of the corresponding skill demand feature entity according to the time difference value;

[0080] In the embodiments of the present application, the time decay coefficient is a value between 0 and 1, which quantifies the timeliness value of information. The coefficient monotonically decreases as the time difference increases.

[0081] In some embodiments of the present application, an exponential decay function is used to calculate the time decay coefficient. Set the half-life parameter λ, the time decay coefficient D(t) = e -λt where t is the time difference. For example, in the field of software development, technology iterates quickly, and λ is set to a large value, so that the coefficient of data more than half a year quickly approaches 0; for basic disciplines, λ is set to a small value. Through this non-linear coefficient mapping, the system can greatly reduce the influence of outdated data on the results, thereby sharply capturing emerging trends in a dynamic environment, for example, a new technology is frequently mentioned in the past month, even if the total frequency is not high, it can also be valued due to its extremely high time decay coefficient (close to 1).

[0082] S203, based on the time decay coefficient and the occurrence frequency, determining the demand weight value of the corresponding skill demand feature entity.

[0083] In the embodiments of the present application, the demand weight value is a comprehensive index for measuring the importance of the node. This step integrates the timeliness factor into the frequency statistics through weighted summation or product operation.

[0084] In some embodiments of the present application, the calculation formula is: W(e) = sum[F(e, d) x D(t_d)], where W(e) is the demand weight value of the skill demand feature entity e, F(e, d) is the occurrence frequency of the skill demand feature entity e in a single data d, and D(t_d) is the time decay coefficient corresponding to the single data d. The system traverses all data entries containing the skill demand feature entity, and accumulates the frequency contribution after time correction. This calculation method overcomes the historical inertia error caused by simply relying on cumulative frequency, so that the weight of the "out of fashion" technology (such as Flash development) which has a large historical total but is not in demand recently is lowered, and the weight of the technology (such as Artificial Intelligence Generated Content (AIGC) application development) which has recently grown explosively is effectively improved.

[0085] As can be seen, by introducing the dynamic decay calculation logic based on the time difference, the frequency data is weighted and corrected using the time decay coefficient, so that the generated feature map can accurately reflect the real skill demand heat of the current time section, effectively eliminating the interference of outdated information, and ensuring that the task chain constructed subsequently has high market sensitivity and timeliness value.

[0086] In some embodiments of the present application, based on the skill demand feature map, a corresponding target task unit is matched from a preset task unit library, including:

[0087] S301, in the skill demand feature map, a target map node is determined;

[0088] In an embodiment of the present application, the target map node refers to a node with a high priority attribute or belonging to a critical path in the generated skill demand feature map, representing the core ability entity that must be mastered to achieve the professional or business goal associated with the target object identifier.

[0089] In some embodiments of the present application, the system traverses all nodes in the skill demand feature map and reads the demand weight value in each node attribute parameter. The system marks the nodes with a demand weight value greater than a preset threshold or the nodes ranked in the top preset proportion according to the demand weight value from high to low as target map nodes. In addition, the system can also exclude those nodes that have been marked as "mastered" although they have high weights, thereby focusing on the skill gap to be improved.

[0090] S302, for each preset task unit in the task unit library, determine the content coverage index, difficulty gradient matching degree, and historical contribution effectiveness coefficient between the preset task unit and the target map node;

[0091] In an embodiment of the present application, the preset task unit is the smallest execution granularity stored in the task unit library, including knowledge point labels, operation difficulty levels, and historical statistical data.

[0092] The content coverage index represents the degree of overlap between the knowledge point set contained in the task unit and the skill semantics represented by the target map node. In some embodiments of the present application, the system uses natural language processing technology to extract the description text feature vector of the preset task unit, calculates the cosine similarity between the feature vector and the target map node feature vector, or calculates the Jaccard similarity coefficient between the metadata label set of the task unit and the sub-concept set associated with the map node as the content coverage index.

[0093] The difficulty gradient matching degree represents the distance between the difficulty setting of the task unit and the required ability level of the target graph node. In some embodiments of the present application, the system obtains the ability level requirement (such as "master" corresponding to level 5) in the target graph node attribute and the difficulty label level (such as level 3) of the preset task unit. The system calculates the matching degree of the two by using the Gaussian distribution function or the reciprocal distance formula. For the preset task unit and the target graph node whose difficulty label level is slightly higher or equal to the ability level requirement (for example, the difference between the difficulty label level and the ability level requirement is greater than zero and less than or equal to a preset value (such as 2)), a larger difficulty gradient matching degree is given to stimulate the potential of the user; and for the preset task unit and the target graph node whose difficulty label level and the ability level requirement differ too much, a smaller difficulty gradient matching degree is given.

[0094] The historical contribution effectiveness coefficient represents the success conversion rate of the task for actually mastering the skill based on the group historical data statistics. In some embodiments of the present application, the system queries the historical log to statistically obtain the pass rate or score improvement amplitude of the historical user who has completed the preset task unit in the subsequent skill evaluation for the target graph node. The greater the improvement amplitude, the higher the historical contribution effectiveness coefficient, thereby quantifying the practical value of the task.

[0095] S303, determining the matching score between the preset task unit and the target graph node based on the content coverage degree index, the difficulty gradient matching degree, and the historical contribution effectiveness coefficient;

[0096] In embodiments of the present application, the matching score is a comprehensive value for ranking the suitability of the task unit.

[0097] In some embodiments of the present application, the system uses a multi-factor linear weighting model for calculation. The system pre-sets the weight coefficients of each index, and performs weighted summation on the normalized content coverage degree index, the difficulty gradient matching degree, and the historical contribution effectiveness coefficient to obtain the matching score. The matching score not only considers "whether the content is correct", but also takes into account "whether the difficulty is appropriate" and "whether the effect is good".

[0098] S304, determining the target task unit from the task unit library according to the matching score.

[0099] In some embodiments of the present application, the system ranks all candidate preset task units for the same target graph node in descending order of the matching score, and selects the top K task units (Top-K) as the target task unit. Alternatively, the system sets a score threshold, and only retains the task units with a score higher than the threshold, thereby filtering out low-quality or irrelevant tasks.

[0100] It can be seen that the embodiments of the present application realize accurate mapping between task units and demand nodes by constructing a multi-dimensional evaluation system including semantic coverage, difficulty adaptation and historical performance. The scheme not only ensures the consistency of the selected tasks in knowledge points, but also ensures the executability and effectiveness of the tasks through the introduction of difficulty matching and historical data verification, thereby generating an execution sequence with high quality and high conversion rate.

[0101] In some embodiments of the present application, the target task unit is determined from the task unit library according to the matching score, comprising:

[0102] S401, determining the target task quantity based on the demand weight value of the target graph node;

[0103] In the embodiments of the present application, the target task quantity refers to the number of execution units that should be included in the final task chain generated for a single target graph node of an individual. The target task quantity can be positively correlated with the demand weight value of the target graph node. Positive correlation means that as the demand weight value increases, the corresponding task allocation quantity also increases. The basis for the design of this logic is that the higher the demand weight value of a node, the higher the importance of the node in the external demand data stream, and therefore more computing resources or training frequencies need to be allocated to ensure that the key feature is fully covered or mastered.

[0104] S402, determining the target task unit matched with the target task quantity from the task unit library according to the matching score.

[0105] In some embodiments of the present application, the system first obtains the matching scores of all candidate preset task units for the target graph node, and constructs a sorting sequence of the candidate preset task units in descending order of the matching scores. Then, the system starts from the top of the sorting sequence and sequentially selects a number of preset task units equal to the target task quantity, and confirms the set of selected units as the final target task unit. If the total number of task units in the candidate library that meet the condition is less than the target task quantity, the system will select all available task units, and a gap log can be generated to prompt the administrator to expand the library. This step ensures that the tasks selected into the final execution sequence not only meet the weight configuration in terms of quantity, but also are the optimal solution in terms of quality in the current library.

[0106] It can be seen that the embodiments of the present application realize the automation and intelligentization of resource allocation by establishing a positive correlation mapping mechanism between task quantity and feature weight. The scheme can dynamically adjust the execution load according to the urgency of market skill demand, automatically increase the training proportion for high-value core skills, and simplify the processing for edge skills, thereby ensuring the construction of core competitiveness while effectively optimizing the overall execution efficiency and time cost.

[0107] In some embodiments of the present application, a target task chain for the target object identifier is generated according to the target task units, including:

[0108] S501, a directed acyclic graph including the target task units is constructed;

[0109] In embodiments of the present application, the directed acyclic graph (DAG) is a graph theory data structure used to represent the non-cyclic dependency relationship between the plurality of target task units. The node in the graph represents a specific target task unit, and the directed edge represents the logical order or pre-condition of task execution.

[0110] In some embodiments of the present application, the system reads the metadata configuration of each target task unit and parses the "predecessor task identifier" or "input parameter source" field defined therein. If the input of task B depends on the output of task A, or task B requires to be started after task A is completed, the system constructs a directed edge from task A to task B. Through this step, the system organizes the discrete plurality of target task units into a structured graph with strict logical timing, effectively preventing the occurrence of dead loop or logical conflict in the task execution flow.

[0111] In embodiments of the present application, in order to ensure that the constructed graph is a directed acyclic graph, the system runs a loop detection algorithm when adding a dependency edge. The specific process is as follows: the system maintains a temporary adjacency list, and before attempting to add an edge from task A to task B, it first detects whether there is a reachable path from task B to task A. If a reachable path is detected, it means that adding the edge will form a closed loop. At this time, the system triggers a conflict resolution mechanism: first, query the "dependency strength attribute" of task A and task B. If the dependency relationship is marked as "weak dependency" or "suggested execution order", the system directly discards the edge and breaks the loop; if both are "strong dependency", the system retains the newer task node based on the task creation timestamp, marks the older task node as "conflict to be manually audited", and temporarily removes it from the graph, or automatically creates a virtual "decoupling intermediate node" to reconstruct the dependency relationship. This processing logic ensures that the generated graph structure always satisfies the topological properties of DAG, avoiding the risk of deadlock.

[0112] S502, historical execution log data of the target object identifier is obtained, and the historical execution log data includes historical completed tasks;

[0113] In the embodiments of the present application, the historical execution log data refers to the detailed records stored in the system database about each task operation performed in the past time period for the target object identifier, including task code, execution timestamp, execution result status, and performance score. The historical completed tasks refer to the task set that is explicitly marked as "execution success", "examination pass", or "accepted" in the log data.

[0114] In some embodiments of the present application, the system initiates a query request to the backend log server through the target object identifier as the index key to retrieve the full amount of historical operation records. The system filters the original log to exclude records of execution failure, abandonment in the middle, or expired results, and only extracts the list of completed task identifiers that are still judged as valid in the current state. This step aims to obtain the personalized ability portrait of the user to provide data support for subsequent differentiated task planning.

[0115] S503, pruning the directed acyclic graph based on the historical completed tasks to obtain a processed directed acyclic graph;

[0116] In the embodiments of the present application, the pruning refers to the operation of dynamically removing redundant nodes or invalid branches from the directed acyclic graph while keeping the original graph structure dependency logic unchanged.

[0117] In some embodiments of the present application, the system traverses all nodes in the directed acyclic graph. For each node, the system judges whether the corresponding task identifier exists in the list of historical completed tasks. If it exists, it means that the task belongs to the mastered or unnecessary repeated execution content for the current user, and the system removes the node from the graph and updates the connection relationship of the associated edge. For example, if the original graph is A→B→C and A has been completed, the pruned graph structure becomes B→C, and at this time the B node becomes a starting node with an in-degree of 0. Through pruning, the system can eliminate the stock skill tasks that the user has mastered well, and realize the individualization of the task graph.

[0118] S504, generating a target task chain for the target object identifier according to the processed directed acyclic graph.

[0119] In some embodiments of the present application, the system performs a topological sorting algorithm on the processed directed acyclic graph to convert the complex graph structure into a linear execution list. For the nodes in the graph that are in parallel relationship (i.e. nodes without direct dependency relationship), the system can perform secondary sorting according to the task weight value, and preferentially arrange the tasks with higher weight values. The generated task chain only contains high-value tasks that the user has not mastered and needs to perform immediately, and constitutes a customized growth path for the user.

[0120] It can be seen that, by constructing a task dependency graph and combining user historical data for dynamic pruning, the embodiments of the present application realize intelligent deduplication and personalized customization of execution paths. The scheme can automatically identify and eliminate historical tasks completed by the user, avoid resource waste caused by repeated training, ensure that the generated task chain always focuses on the user's ability gap, and significantly improve the accuracy and execution efficiency of task planning.

[0121] In some embodiments of the present application, according to the processed directed acyclic graph, a target task chain for the target object identifier is generated, including:

[0122] S601, obtaining the estimated execution duration of each node in the processed directed acyclic graph;

[0123] In an embodiment of the present application, the estimated execution duration refers to the average time cost or standard working hours required to complete the target task unit corresponding to the node. This duration not only reflects the complexity of the task itself, but also includes the time overhead of reading documents, writing code, and debugging and correcting necessary processes.

[0124] In an embodiment of the present application, for a brand new task unit lacking historical execution data, the system determines the estimated execution duration using a cold start estimation model based on attribute regression. Specifically, the system pre-constructs a duration regression model, which takes the "text length" (e.g., reading volume), "code line / interaction step number" (representing operation amount), "knowledge point complexity coefficient" (representing thinking density), and "task type" (such as single selection, question and answer, and practical operation) of the task as input features. When a new task is introduced, the system extracts the above feature inputs into the regression model and outputs a baseline estimated duration. For example, for a Python practice task containing 500 words of reading material and 10 lines of code fill-in-the-blank, the model calculates a baseline duration of 15 minutes according to the fitting curve of historical similar tasks. In addition, the system also introduces "similar task transfer" logic to find the highest similarity stored task (similarity>0.9) in the library with the new task semantics, and directly reuses the historical average duration of the similar task as the estimated execution duration of the new task.

[0125] In some embodiments of the present application, the system reads the basic duration label of the task from the metadata of the preset task unit library. Further, in order to improve the accuracy of the estimate, the system calls the historical user execution database to calculate the median or weighted average of the time duration consumed by all historical users to complete the task. In addition, the system can also combine the personal historical efficiency factor of the target object identifier (for example, the user's past speed of completing similar difficulty tasks is 20% faster than the average level) to make personalized corrections to the basic duration, thereby obtaining the estimated execution duration for the specific user.

[0126] S602, mapping the estimated execution duration to a weight value of a corresponding node in the directed acyclic graph;

[0127] In embodiments of the present application, the weight value specifically refers to a "cost" or "resistance" parameter in the path planning algorithm. In the graph model, the weight is usually assigned to the edge, and in this step, the system converts the attribute (duration) of the node into the cumulative cost in the graph traversal process.

[0128] In some embodiments of the present application, the mapping process adopts a direct linear mapping, that is, the weight value is equal to the estimated execution duration (for example, 2 hours is mapped to a weight of 2.0). In more complex scenarios, the system can introduce a time urgency adjustment factor. If the target object identifier sets a shorter overall deadline, a non-linear high weight penalty is given to the task with longer duration to force the algorithm to avoid paths that take too long. In order to adapt to the standard shortest path algorithm, the system can logically transfer the weight value of the node to all the incoming edges pointing to the node, or include the inherent weight of the node when calculating the total weight of the path.

[0129] S603, using Dijkstra algorithm to determine a target path sequence from the start node to the end node in the directed acyclic graph, with the optimization objective of minimizing the sum of weight values on the path;

[0130] In embodiments of the present application, the target path sequence refers to an ordered set of nodes in the directed acyclic graph that are passed through from the start node representing the current state to the end node representing the final capability target. The optimization objective is to minimize the total execution duration of all tasks in the sequence, that is, to find the "fastest path to achieve". This step is particularly suitable for scenarios where there are parallel and mutually exclusive branches in the graph (for example, technical solution A path or technical solution B path can be selected to achieve the same target).

[0131] In some embodiments of the present application, the system initializes the distance of all nodes to the start node as infinity, and the distance of the start node as 0. Using the greedy strategy of Dijkstra algorithm, the system maintains a priority queue, and selects the node with the smallest cumulative weight value each time to perform relaxation operation (Relaxation) to update the cumulative weight value of its adjacent nodes. Since it is a directed acyclic graph, this algorithm can guarantee convergence within a finite number of steps. Finally, the system constructs an optimal path with the shortest total time consumption by backtracking the recorded predecessor nodes. This path represents the execution scheme that the user can achieve the target fastest while meeting all logical dependencies.

[0132] S604, generating a target task chain for the target object identifier based on the target path sequence.

[0133] In some embodiments of this application, the system extracts the task identifiers of all nodes in the target path sequence and arranges them into a linked list structure according to their topological order in the path. The system can also attach the estimated execution time to each task node to generate a project schedule with estimated time progress. For other branch nodes not selected in the target path sequence, the system marks them as "optional supplementary tasks" and does not treat them as core deliverables, thereby focusing the user's execution efforts.

[0134] As can be seen, this application's embodiments introduce a time dimension as the core cost function and utilize the shortest path algorithm to automatically search for the optimal solution in complex task dependency networks, achieving task chain construction oriented towards time efficiency. This scheme can automatically plan the execution path with the shortest total time and highest efficiency when multiple technical routes or execution branches exist, reducing users' time costs and maximizing execution benefits.

[0135] In some embodiments of this application, after generating a target task chain for a target object identifier based on the target task unit, the method further includes:

[0136] S701, Receive execution feedback data for the current task execution node in the target task chain, the execution feedback data including the task execution completion score;

[0137] In the embodiments of this application, the current task execution node refers to the task unit that the target object is processing or has just submitted a completion result in the target task chain. Execution feedback data is a data packet transmitted back to the server through the user terminal or automated evaluation system, used to characterize the specific situation of the task execution. The task execution completion score is a quantitative indicator of the quality of the execution result, usually a standardized value (such as 0 to 100 points or a coefficient of 0 to 1.0).

[0138] In some embodiments of this application, when the current task is a programming code task, the system runs the submitted code in a backend sandbox environment and calculates the task completion score based on the pass rate of the test cases; when the current task is a knowledge question-and-answer task, the system directly generates a score based on the accuracy rate. This data reflects the target object's mastery of the current knowledge point in real time, providing an objective basis for the dynamic adjustment of subsequent links.

[0139] S702. Based on the task completion score, perform task execution node insertion or deletion processing on the target task chain.

[0140] In the embodiments of this application, this step aims to construct a closed-loop adaptive adjustment logic to dynamically correct the pre-generated static plan based on real-time capability feedback, so as to make the target task chain more accurate and more timely.

[0141] It can be seen that, by introducing the dynamic link reconstruction logic based on real-time feedback, the embodiments of the present application realize the adaptive matching of the task chain to the actual ability of the user. The scheme can automatically insert a tutoring task to reduce the gradient when the user performs poorly, and automatically remove redundant tasks to improve efficiency when the user performs excellently, thereby converting a static plan into a dynamic personalized guidance path, ensuring the smoothness and efficiency of the execution process.

[0142] In some embodiments of the present application, based on the task execution completion score, the target task chain is subjected to task execution node insertion processing or task execution node deletion processing, including at least one of the following steps:

[0143] S801, if the execution completion score is less than a preset first threshold, obtaining a to-be-supplemented task unit matched with the current task execution node in a task unit library, and based on the to-be-supplemented task unit, performing task execution node insertion processing on the target task chain;

[0144] In the embodiments of the present application, the first threshold is a critical value (for example, 60 points or the passing line) for determining whether the target object passes the current examination or meets the minimum mastery requirement. The to-be-supplemented task unit refers to a backup task entity used for remedial teaching, basic reinforcement, or fine-grained decomposition of the current unmastered content.

[0145] In some embodiments of the present application, when the system detects that the score is lower than the first threshold, a remedial strategy is triggered. The system first analyzes the knowledge point label associated with the current task execution node, and then searches for a task unit with the same knowledge point label but a lower difficulty level than the current task in the task unit library; or searches for a basic concept task unit having a “pre-depending” relationship with the current task. The system inserts one or more to-be-supplemented task units retrieved into the target task chain in a logical progressive order immediately after the current node, and updates the pointer connection relationship of each node in the linked list. This process ensures that when the target object fails to pass the current checkpoint, it is not mechanically repeated with the same task, but through lower-gradient auxiliary tasks to consolidate the foundation, preventing task chain blockage due to too large difficulty span.

[0146] In the embodiments of the present application, a multi-dimensional knowledge graph is constructed at the bottom of the task unit library. The entities in the graph include not only task units, but also "knowledge point" entities. The entities are connected by semantic edges such as "contains", "precedes", "is-a", etc. The specific retrieval logic is as follows: when insertion processing is needed, the system first locates the "knowledge point identity (ID)" mounted by the current task execution node; then, it traverses along the reverse path of the "is-a" edge (i.e., in the direction of concretization or basification) or along the reverse path of the "precedes" edge (i.e., in the direction of preceding knowledge) in the graph. For example, if the current task is associated with the "double-linked list reversal" knowledge point, the system queries the graph and finds that the preceding knowledge points are "single-linked list traversal" and "pointer basics". The system then obtains all task sets with a difficulty level of Level-1 mounted under the "pointer basics" knowledge point, and sorts them according to the user's historical preferences (such as whether to prefer video or practical operation), and selects the first ranked task as the task unit to be supplemented. This retrieval method based on the topology of the graph ensures the logical relevance and effectiveness of the remedial task and the current obstacle point.

[0147] S802, if the execution completion degree score is greater than a preset second threshold value, in the target task chain, a to-be-skipped task execution node located after the current task execution node and having a task difficulty level less than a preset level threshold value is determined, and the target task chain is subjected to task execution node deletion processing based on the to-be-skipped task execution node, the second threshold value being greater than or equal to the first threshold value.

[0148] In the embodiments of the present application, the second threshold value is a high standard value (e.g., 90 points or excellent line) for determining whether the target object has perfectly mastered or exceeded the current task. The to-be-skipped task execution node refers to a low-value task that is too simple, repetitive or ineffective for the target object who has already shown a high level of ability in the originally planned subsequent path. The preset level threshold value is usually set based on the difficulty level of the task that has been completed (for example, if the task that has been completed has a difficulty level of 5, then tasks with difficulty levels of 3 and below can be considered as low-level).

[0149] In some embodiments of this application, when the completion score indicates that the target object has perfectly mastered or exceeded the current task, the system traverses the subsequent nodes in the target task chain that have not yet been executed. The system reads the difficulty attribute of each subsequent node and marks those nodes involving the same or subordinate knowledge points and whose difficulty is significantly lower than the currently completed task as task execution nodes to be skipped. Subsequently, the system performs a deletion operation, removing these nodes from the linked list structure and directly pointing the predecessor node of the deleted node to its successor node. This step accelerates and optimizes the execution path by dynamically pruning inefficient repetitive practice.

[0150] As can be seen, this embodiment of the application automatically inserts downgrade tutoring tasks when scores are low and automatically removes low-level redundant tasks when scores are high. This scheme enables the task chain to scale and adapt according to the user's real-time performance. This not only effectively solves the problem of users experiencing lag due to weak foundations, but also avoids the time wasted by high-level users due to repetitive low-difficulty practice, maximizing the output of ability improvement per unit of time.

[0151] In some embodiments of this application, matching corresponding target task units from a preset task unit library based on a skill requirement feature map includes:

[0152] S901. For each target graph node in the skill requirement feature graph, construct a task candidate set based on the intersection operation of knowledge point labels;

[0153] In embodiments of this application, a target graph node is defined as a skill atom, which is the smallest indivisible semantic unit of capability parsed from an external demand data stream (e.g., "Python list comprehension" rather than the general "Python programming"). Knowledge point tags are standardized terms pre-annotated in the task unit metadata, used to characterize the specific technical theories or operational points covered by the task.

[0154] In some embodiments of this application, the system employs an inverted index technique to perform intersection operations. The system pre-constructs an index structure with knowledge point tags as keys and task unit identifier lists as values. When processing a target graph node, the system extracts the set of knowledge points contained in that node, retrieves all task units with non-empty intersections with it from the index structure, and adds them to the task candidate set. This step, through intersection logic in set theory, filters out tasks with extremely low relevance before performing complex weight calculations, reducing the search space for subsequent calculations and lowering the system load.

[0155] S902. Construct a weighted bipartite graph connecting the target graph nodes with each task unit in the task candidate set;

[0156] In the embodiments of this application, a weighted bipartite graph is a special graph structure in which the set of nodes is divided into two disjoint subsets: the first subset contains all target graph nodes (i.e., the demand side), and the second subset contains all preset task units (i.e., the supply side) in the task candidate set. Edges in the graph exist only between the first and second subsets; there are no connections within the subsets themselves. In the step of constructing the weighted bipartite graph, the matching score from the above embodiments can be assigned to the corresponding edge connecting nodes in the first and second subsets.

[0157] S903. Solve the weighted bipartite graph using a preset matching algorithm, determine the target task matching relationship, and identify the matched task unit as the target task unit.

[0158] In the embodiments of this application, the preset matching algorithm refers to a graph optimization algorithm that maximizes the sum of the weights of the selected edges under certain constraints, such as a variant of the Kuhn-Munkres Algorithm (KM) or the Minimum Cost Maximum Flow Algorithm.

[0159] In some embodiments of this application, the system runs the KM algorithm with the global optimization objective of maximizing the sum of the weights of all matching edges. This algorithm, by introducing a labeling mechanism, finds a complete or optimal match in polynomial time. Compared to simple Top-K sorting, weighted bipartite graph matching can solve the problems of "resource competition" and "global optimum." For example, when a high-value task can train both skill A and skill B, but can only occur once in the current path planning, the algorithm can allocate it to the side with the greater benefit based on its global weight contribution, or select a set of highly complementary task combinations to cover all skill atoms, avoiding suboptimal overall efficiency caused by local greed. The system ultimately determines the set of task units pointed to by the matching edges output by the algorithm as the target task unit.

[0160] As can be seen, the embodiments of this application introduce a candidate set pre-screening mechanism based on inverted index and a global matching algorithm based on weighted bipartite graph. The pre-screening eliminates irrelevant tasks to improve computational performance, and the bipartite graph matching algorithm optimizes the correspondence between skills and tasks in the global dimension. By maximizing the overall matching weight, the suboptimal resource allocation problem that may be caused by local greedy strategies is solved, and the mathematically optimal solution of the constructed task chain in terms of teaching effectiveness and knowledge coverage is achieved.

[0161] In some embodiments of this application, determining the number of target tasks based on the demand weight values ​​of target graph nodes includes:

[0162] S1001. Based on the required weight values ​​of the target map nodes, determine the preset mastery confidence threshold of the target map nodes through a nonlinear mapping function;

[0163] In the embodiments of this application, the preset mastery confidence threshold refers to the minimum probability value required for the system to determine that the target object has reached a level of mastery of the skill or knowledge point represented by the node that meets market expectations, and the value range is usually [0, 1]. This preset mastery confidence threshold is non-linearly positively correlated with the demand weight value.

[0164] In some embodiments of this application, the system employs a variant of the Sigmoid function as the nonlinear mapping function. The system sets a base confidence level (e.g., 0.5), and as the demand weight value increases, the preset mastery confidence threshold grows in an "S"-shaped curve, approaching 1.0. For example, for core high-frequency test points with a demand weight value of 0.9, the system-calculated confidence threshold might be 0.98 (i.e., mastery is required); while for peripheral knowledge points with a demand weight value of 0.3, the confidence threshold might only be 0.60 (i.e., basic understanding is required). This mapping logic reflects a differentiated configuration strategy of "high-value skills require high-precision mastery, while low-value skills only require general knowledge."

[0165] S1002. Obtain the knowledge complexity coefficient of the target graph node, and calculate the total value of the target training impulse based on the preset mastery confidence threshold and knowledge complexity coefficient.

[0166] In embodiments of this application, the knowledge complexity coefficient is a quantitative indicator pre-stored in the graph metadata, used to characterize the amount of cognitive resources required to master the knowledge point (e.g., the coefficient for mastering "loop structure" is 1.2, and the coefficient for mastering "dynamic programming algorithm" is 5.0). The target training impulse total value refers to the total amount of task stimuli theoretically required to overcome the cognitive resistance of the knowledge point and reach the preset mastery confidence level.

[0167] In some embodiments of this application, the formula for calculating the total target training impulse is: I target =C k ×ln(P thresh / (1-P thresh Among them, It arget C represents the total training impulse value. k Let P be the knowledge complexity coefficient. threshA preset mastery confidence threshold is set. The formula is based on the inverse operation of the logistic model in the item response theory (IRT), which means that as the target confidence increases, the required training impulse will rise exponentially, rather than simply linearly.

[0168] S1003, obtain the average effective cognitive gain value of the candidate task units in the task unit library, divide the target training impulse total value by the average effective cognitive gain value, and round up to obtain the target task quantity.

[0169] In embodiments of the present application, the average effective cognitive gain value refers to the average ability improvement amount or impulse contribution value that a single standard task unit can bring to the target object after being executed. This value is obtained by the system based on Bayesian inference of historical big data.

[0170] In some embodiments of the present application, the system first selects the top N (TOP-N) candidate tasks for the target graph node from the library, calculates the historical average gain of these tasks. Then, the division operation is performed: Q = ⌈Itarget / (G avg × δ)⌉. Wherein, Q is the target task quantity, ⌈⌉ represents the rounding up function, G avg is the average effective cognitive gain value, and δ is a redundancy safety factor (for example, 1.2, used to offset the forgetting curve effect). Through this step, the abstract "weight" is converted into the specific "question quantity" or "project operation times" calculated by the system, avoiding the problem of not being firm due to too few tasks or marginal utility diminishing due to too many tasks.

[0171] As can be seen, embodiments of the present application propose a double constraint quantitative model based on cognitive load and mastery confidence. Through nonlinear mapping, differentiated mastery standards are determined, the required physical training impulse is calculated in combination with the complexity of the knowledge points, and finally converted into a specific task quantity, achieving a balance between "market value" and "learning cost" in the order of magnitude of the task, ensuring that core difficulties have enough training support, and simple knowledge points do not occupy too many time resources.

[0172] In a second aspect, embodiments of the present application provide a task chain construction system for a skill training platform. The task chain construction system for the skill training platform is configured to execute the task chain construction method for the skill training platform as described in any of the above embodiments.

[0173] In a third aspect, embodiments of the present application provide a computer readable storage medium storing a computer program configured to be executed by a processor to implement the task chain construction method for the skill training platform as described in any of the above embodiments.

[0174] The above has carried out the detailed introduction to the embodiment of the application, the principle and implementation mode of the application have been set forth by applying specific examples in this paper, the above embodiment explanation is only for helping understanding the method of the application and its core idea; at the same time, for the person skilled in the art, according to the idea of the application, there will be changes in specific implementation mode and application range, and the above-mentioned, the content of the specification should not be understood as the limitation of the application.

Claims

1. A method for constructing a task chain for a skill-oriented practical training platform, characterized in that, The task chain construction method for the skill training platform comprises the following steps: Obtain external demand data flow associated with a target object identifier in a skill training platform, wherein the external demand data flow comprises a plurality of skill demand description texts with time sequence marks; Perform entity extraction processing on the external demand data flow to obtain a plurality of skill demand characteristic entities; Determine a demand weight value of each skill demand characteristic entity according to the occurrence frequency of the skill demand characteristic entity in the external demand data flow and the time sequence mark; Take the skill demand characteristic entity as a graph node and write the demand weight value into an attribute parameter of the corresponding graph node to obtain a skill demand characteristic graph; Match a corresponding target task unit from a preset task unit library based on the skill demand characteristic graph; Generate a target task chain for the target object identifier according to the target task unit. 2.The method of claim 1, wherein, The method for determining the demand weight value of each skill demand characteristic entity according to the occurrence frequency of the skill demand characteristic entity in the external demand data flow and the time sequence mark comprises the following steps: Determine a time difference value between the time sequence mark and the current system time; Determine a time decay coefficient of the corresponding skill demand characteristic entity according to the time difference value; Determine the demand weight value of the corresponding skill demand characteristic entity based on the time decay coefficient and the occurrence frequency. 3.The method of claim 1, wherein, The method for matching the corresponding target task unit from the preset task unit library based on the skill demand characteristic graph comprises the following steps: Determine a target graph node in the skill demand characteristic graph; Determine a content coverage index, a difficulty gradient matching degree and a historical contribution effectiveness coefficient between each preset task unit in the task unit library and the target graph node; Determine a matching score between the preset task unit and the target graph node based on the content coverage index, the difficulty gradient matching degree and the historical contribution effectiveness coefficient; Determine the target task unit from the task unit library according to the matching score.

4. The method of claim 3, wherein the method further comprises: The method for determining the target task unit from the task unit library according to the matching score comprises the following steps: Determine a target task quantity based on the demand weight value of the target graph node; Determine the target task unit matching the target task quantity from the task unit library according to the matching score.

5. The method of claim 1, wherein the method further comprises: The method for generating a target task chain for the target object identifier according to the target task unit comprises the following steps: Construct a directed acyclic graph comprising the target task unit; Obtain historical execution log data of the target object identifier, wherein the historical execution log data comprises historical completed tasks; Perform pruning processing on the directed acyclic graph based on the historical completed tasks to obtain a processed directed acyclic graph; Generate a target task chain for the target object identifier according to the processed directed acyclic graph.

6. The method of claim 5, wherein the method further comprises: The method for generating a target task chain for the target object identifier according to the processed directed acyclic graph comprises the following steps: Obtain an estimated execution duration of each node in the processed directed acyclic graph; mapping the estimated execution duration as a weight value of a corresponding node in the directed acyclic graph; minimizing the sum of the weight values on the path as an optimization objective, using Dijkstra algorithm to determine a target path sequence from a starting node to a terminal node in the directed acyclic graph; generating a target task chain for the target object identifier based on the target path sequence.

7. The method of claim 1, wherein the method further comprises: receiving a task chain from the task chain database; and displaying the task chain on the display device. After the step of generating the target task chain for the target object identifier based on the target task unit, the method further includes: receiving execution feedback data for a current task execution node in the target task chain, the execution feedback data including a task execution completion degree score; based on the task execution completion degree score, performing task execution node insertion processing or task execution node deletion processing on the target task chain.

8. The method of claim 7, wherein the method further comprises: The step of performing task execution node insertion processing or task execution node deletion processing on the target task chain based on the task execution completion degree score includes at least one of the following steps: if the execution completion degree score is less than a preset first threshold, obtaining a to-be-supplemented task unit matched with the current task execution node in the task unit library, and performing task execution node insertion processing on the target task chain based on the to-be-supplemented task unit; if the execution completion degree score is greater than a preset second threshold, determining a to-be-skipped task execution node located after the current task execution node in the target task chain and having a task difficulty level less than a preset level threshold, and performing task execution node deletion processing on the target task chain based on the to-be-skipped task execution node, the second threshold being greater than or equal to the first threshold. 9.A task chain construction system for a skill-oriented practical training platform, characterized in that, The task chain construction system for the skill training platform is configured to perform the task chain construction method for the skill training platform according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program configured to be executed by the processor to implement the task chain construction method for the skill training platform according to any one of claims 1 to 8. The computer readable storage medium stores a computer program configured to be executed by the processor to implement the task chain construction method for the skill training platform according to any one of claims 1 to 8.

Images