Task adaptive assignment in intelligent meeting collaboration and system

Abstract

The present application relates to the technical field of intelligent conference collaboration, and especially relates to a task adaptive allocation method and system in intelligent conference collaboration. A knowledge matching network is constructed based on task knowledge dependency and participant knowledge graph, and a knowledge propagation path is deduced, necessary knowledge nodes and gap nodes are identified, and a knowledge complementarity degree is calculated. According to the analysis result and the collaboration granularity characteristics, a task set is reconstructed through a bidirectional adjustment mechanism, tasks are decomposed and a knowledge transmission interface is established, or tasks are aggregated and a knowledge sharing mechanism is established. According to the reconstructed task structure and the cognitive switching cost, a cognitive state transition graph is constructed, a task execution sequence with minimum cognitive switching cost is planned, a time window is aligned, and a task allocation instruction is generated. The present application improves the adaptability and execution efficiency of task allocation, and reduces the cognitive switching cost.

Description

Technical Field

[0001] This invention relates to the field of intelligent meeting collaboration technology, and in particular to a method and system for adaptive task allocation in intelligent meeting collaboration. Background Technology

[0002] In intelligent meeting collaboration scenarios, task allocation is a crucial factor determining collaboration efficiency. Current conventional practices typically rely on participants' roles, positions, or predefined skill tags for static task assignment. The system pre-binds tasks to personnel based on the meeting agenda, for example, by setting a fixed execution sequence through a workflow engine or using a polling mechanism to allocate tasks sequentially. Some solutions introduce task dependency graphs, but only focus on the order of tasks, failing to delve into the connections at the knowledge level.

[0003] Conventional approaches suffer from two main drawbacks. First, static assignment based on roles or tags ignores the knowledge dependencies required for task execution. Participants can be assigned to tasks outside their own knowledge base, leading to missing key knowledge nodes and requiring repeated communication or task transfer during collaboration, thus reducing overall efficiency. Second, pre-defined task sequences lack explicit management of knowledge transfer interfaces. Participants must re-understand the context when switching tasks, incurring high cognitive switching costs. This is especially problematic in multi-task parallel or knowledge-intensive meetings, where frequent switching between different knowledge domains disrupts attentional continuity and impacts execution quality. Furthermore, existing methods cannot dynamically identify knowledge gaps, and task decomposition and aggregation lack objective criteria, making it difficult to adapt to changes in participants' knowledge levels or task requirements during meetings. This results in assignments that are out of sync with real-time collaboration needs. Summary of the Invention

[0004] This invention provides a method and system for adaptive task allocation in intelligent meeting collaboration, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides a task adaptive allocation method in intelligent meeting collaboration, comprising:

[0006] Based on the knowledge coverage relationship between the task knowledge dependency characteristics of the meeting and the knowledge graph characteristics of the meeting participants, a knowledge matching network is constructed and the knowledge propagation path is deduced. The necessary knowledge nodes and knowledge gap nodes for task execution are identified, the knowledge complementarity index among participants is calculated, and the knowledge dependency analysis results are obtained.

[0007] Based on the knowledge dependency parsing results and task collaboration granularity characteristics, the task set is reconstructed in terms of collaboration granularity through a two-way adjustment mechanism of task decomposition and aggregation. When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks and a knowledge transfer interface between sub-tasks is established according to the dependency order of the necessary knowledge nodes. When the overlap of the necessary knowledge nodes of multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, multiple tasks are aggregated into a collaborative task group and a knowledge sharing mechanism is established to obtain the task structure after collaboration granularity reconstruction.

[0008] Based on the reconstructed task structure at the collaborative granularity and the cognitive switching cost characteristics of the participants, a cognitive state transition diagram is constructed. Based on the knowledge transfer interface, a task execution sequence with the minimum cognitive switching cost is planned for each participant in the cognitive state transition diagram. The task execution sequence is aligned with the time window based on the participants' historical task execution efficiency characteristics, and a task allocation instruction is generated.

[0009] Based on the knowledge coverage relationship between the task knowledge dependency characteristics of the meeting and the knowledge graph characteristics of the participants, a knowledge matching network is constructed and the knowledge propagation path is deduced. Essential knowledge nodes and knowledge gap nodes for task execution are identified, and the knowledge complementarity index among participants is calculated. The knowledge dependency resolution results include:

[0010] The knowledge items in the task knowledge dependency features are represented as task knowledge nodes, and the knowledge items in the knowledge graph features are represented as participant knowledge nodes. Based on the semantic similarity and knowledge hierarchy relationship between the task knowledge nodes and the participant knowledge nodes, the knowledge matching network containing a set of task knowledge nodes, a set of participant knowledge nodes, and cross-set association edges is constructed.

[0011] In the knowledge matching network, the task knowledge node is selected as the starting node. The reachable participant knowledge nodes are traversed by multi-hop propagation along the associated edges. The knowledge propagation path from the task knowledge node to the participant knowledge node is identified. The knowledge reachability metric of the task knowledge node is calculated based on the path length of the knowledge propagation path and the weight of the associated edges on the path.

[0012] Task knowledge nodes whose knowledge accessibility metric exceeds a preset accessibility threshold are marked as required knowledge nodes, and the remaining task knowledge nodes are marked as knowledge gap nodes;

[0013] Extract a subset of participant knowledge nodes covering the knowledge gap node from the knowledge matching network and obtain the knowledge association strength between participants from it. Calculate the knowledge complementarity index of participants on the knowledge gap node based on the knowledge association strength.

[0014] The knowledge gap nodes are combined with the knowledge complementarity index to form the knowledge dependency analysis result.

[0015] When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks, and a knowledge transfer interface is established between the sub-tasks according to the dependency order of the required knowledge nodes, including:

[0016] When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the set of necessary knowledge nodes corresponding to the current task is extracted. A necessary knowledge node dependency graph is constructed based on the predecessor and successor relationships between knowledge nodes in the set of necessary knowledge nodes. Knowledge nodes without predecessor nodes are identified as dependency starting points and knowledge nodes without successor nodes are identified as dependency ending points in the necessary knowledge node dependency graph. The dependency order of necessary knowledge nodes is determined by traversing the path from the dependency starting point to the dependency ending point.

[0017] The set of necessary knowledge nodes is sequentially segmented according to the dependency order of the necessary knowledge nodes. The necessary knowledge nodes in each segment are combined with the knowledge gap nodes that cover the necessary knowledge nodes in the segment to form a subtask knowledge requirement unit. The current task is decomposed into multiple subtasks based on the subtask knowledge requirement unit.

[0018] For adjacent segments in the required knowledge node dependency graph that have predecessor-successor relationships, the mapping relationship between the required knowledge node identifiers output by the preceding subtask and the required knowledge node identifiers depended on by the subsequent subtask is recorded, and a knowledge transfer interface is established.

[0019] When the overlap of required knowledge nodes in multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, the multiple tasks are aggregated into a collaborative task group and a knowledge-sharing mechanism is established, including:

[0020] Select any two tasks from the set of tasks to be assigned as a task pair, extract the sets of necessary knowledge nodes corresponding to the two tasks respectively, and calculate the overlap of necessary knowledge nodes of the task pair based on the ratio of the number of knowledge nodes in the intersection of the two sets of necessary knowledge nodes to the number of knowledge nodes in the union.

[0021] Traverse all task pairs in the set of tasks to be assigned, and filter out task pairs whose required knowledge node overlap exceeds a preset overlap threshold as candidate aggregation task pairs;

[0022] Extract the knowledge gap node sets corresponding to the two tasks in the candidate aggregation task pair, identify the previous participant subset that can cover the knowledge gap node set of the previous task and the next participant subset that can cover the knowledge gap node set of the next task from the participant set, and calculate the knowledge complementarity index between the previous participant subset and the next participant subset.

[0023] When the knowledge complementarity index exceeds the preset complementarity threshold, the two tasks in the candidate aggregation task pair are aggregated into a collaborative task group.

[0024] Extract the intersection knowledge nodes corresponding to the collaborative task groups as shared knowledge nodes, establish an access mapping relationship between the shared knowledge nodes and the participant knowledge nodes, and establish a knowledge sharing mechanism based on the access mapping relationship.

[0025] Based on the task structure reconstructed at the collaborative granularity and the cognitive switching cost characteristics of the participants, a cognitive state transition diagram is constructed, including:

[0026] Extract all subtask nodes and collaborative task group nodes from the task structure after the collaborative granularity reconstruction, and use the subtask nodes and collaborative task group nodes as cognitive state nodes.

[0027] For each cognitive state node, extract the set of required knowledge nodes and the set of knowledge gap nodes corresponding to the cognitive state node;

[0028] For all cognitive state nodes, calculate the knowledge difference between the set of necessary knowledge nodes corresponding to the previous cognitive state node and the set of necessary knowledge nodes corresponding to the next cognitive state node, and identify a subset of participants from the participant set that can simultaneously cover the set of knowledge gap nodes of the previous cognitive state node and the set of knowledge gap nodes of the next cognitive state node.

[0029] For each participant in the subset of participants, the cognitive switching cost feature of the participant is extracted, and the cognitive switching cost from the previous cognitive state node to the next cognitive state node is calculated based on the knowledge difference and the cognitive switching cost feature.

[0030] A directed edge is established between the previous cognitive state node and the next cognitive state node. The cognitive switching cost is used as the edge weight of the directed edge. A cognitive state transition graph is constructed based on all cognitive state nodes and the directed edge.

[0031] Based on the knowledge transfer interface, the task execution sequence with minimum cognitive switching cost planned for each participant in the cognitive state transition diagram includes:

[0032] For each participant, extract all cognitive state nodes that the participant can cover the set of knowledge gap nodes from the cognitive state transition graph as the reachable cognitive state node set;

[0033] Select an initial cognitive state node from the set of reachable cognitive state nodes. Starting from the initial cognitive state node, traverse the directed edges connecting the nodes in the set of reachable cognitive state nodes in the cognitive state transition graph and extract the edge weights. Accumulate the edge weights of all directed edges on the path from the initial cognitive state node to the target cognitive state node to obtain the total path cost.

[0034] For all cognitive state nodes in the reachable cognitive state node set, the path with the minimum total cost covering all cognitive state nodes in the reachable cognitive state node set is calculated as a candidate task execution sequence. Based on the mapping relationship between knowledge transfer source nodes and knowledge transfer target nodes recorded in the knowledge transfer interface, the cognitive switching cost between adjacent cognitive state nodes in the candidate task execution sequence is adjusted. The candidate task execution sequence with the minimum adjusted total cost is determined as the task execution sequence of the participant.

[0035] Based on the participants' historical task execution efficiency characteristics, the task execution sequence is aligned with a time window to generate task allocation instructions, including:

[0036] Extract the set of necessary knowledge nodes and task complexity features corresponding to each cognitive state node from the task execution sequence, and predict the expected execution time of the cognitive state node based on the set of necessary knowledge nodes, the task complexity features and the historical task execution efficiency features;

[0037] The cognitive switching cost between adjacent cognitive state nodes in the task execution sequence is converted into a time buffer. The start time of the available time window is used as the time starting point, and execution time periods are allocated to the cognitive state nodes in the task execution sequence in sequence within the duration of the available time window.

[0038] When the execution time of a cognitive state node in the task execution sequence exceeds the duration of the available time window, a subset of cognitive state nodes that can be delayed in execution is identified from the task execution sequence, and the subset of cognitive state nodes that can be delayed in execution is mapped to the next available time window of the participant.

[0039] Based on the task execution sequence, the time period corresponding to each cognitive state node and the time window mapping relationship between the subset of cognitive state nodes that can be delayed, task allocation instructions are generated for the meeting participants.

[0040] A second aspect of the present invention provides a task adaptive allocation system for intelligent meeting collaboration, comprising:

[0041] The dependency resolution unit is used to construct a knowledge matching network and deduce the knowledge propagation path based on the knowledge coverage relationship between the task knowledge dependency features of the meeting and the knowledge graph features of the meeting participants. It identifies the necessary knowledge nodes and knowledge gap nodes for task execution, calculates the knowledge complementarity index among participants, and obtains the knowledge dependency resolution results.

[0042] The granularity reconstruction unit is used to reconstruct the collaborative granularity of the task set based on the knowledge dependency parsing results and task collaboration granularity characteristics through a two-way adjustment mechanism of task decomposition and aggregation. When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks and a knowledge transfer interface between sub-tasks is established according to the dependency order of the necessary knowledge nodes.

[0043] The task aggregation unit is used to aggregate multiple tasks into a collaborative task group and establish a knowledge sharing mechanism when the overlap of the required knowledge nodes of multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, thereby obtaining the task structure after the collaborative granularity reconstruction.

[0044] The task allocation unit is used to construct a cognitive state transition diagram based on the task structure reconstructed by the collaborative granularity and the cognitive switching cost characteristics of the participants, plan a task execution sequence with minimum cognitive switching cost for each participant in the cognitive state transition diagram based on the knowledge transfer interface, align the task execution sequence with a time window based on the historical task execution efficiency characteristics of the participants, and generate a task allocation instruction.

[0045] A third aspect of the present invention provides an electronic device, comprising:

[0046] processor;

[0047] Memory used to store processor-executable instructions;

[0048] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0049] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0050] The beneficial effects of this application are as follows:

[0051] Based on the knowledge coverage relationship between the task knowledge dependency characteristics of the meeting and the knowledge graph characteristics of the participants, a knowledge matching network is constructed and the knowledge propagation path is deduced. This can accurately identify the necessary knowledge nodes and knowledge gap nodes for task execution. By calculating the knowledge complementarity index among participants, the efficient matching of tasks and participants' knowledge capabilities is achieved, avoiding execution deviations caused by knowledge blind spots and significantly improving the accuracy of task allocation and collaboration efficiency.

[0052] Based on the knowledge dependency analysis results and task collaboration granularity characteristics, the task set is reconstructed in terms of collaboration granularity through a two-way adjustment mechanism of task decomposition and aggregation. When the knowledge gap node exceeds the coverage capacity of a single participant, the task is automatically decomposed into multiple sub-tasks and a knowledge transfer interface is established between sub-tasks to ensure the executability of complex tasks. When the knowledge nodes of multiple tasks overlap and the participants are highly complementary, the tasks are aggregated into collaborative task groups and a sharing mechanism is established to eliminate duplication of work and information silos, thereby optimizing resource utilization and improving the overall task completion quality.

[0053] Based on the reconstructed task structure at the collaborative granularity and the cognitive switching cost characteristics of participants, a cognitive state transition diagram is constructed. A task execution sequence with minimum cognitive switching cost is planned for each participant. The sequence is then aligned with a time window based on historical task execution efficiency characteristics to generate task allocation instructions. This method effectively reduces the cognitive load and switching costs for participants across multiple tasks, improves task execution fluency and efficiency, ensures high-quality task completion within a given time window, and achieves intelligent adaptive allocation of collaborative meeting tasks. Attached Figure Description

[0054] Figure 1 A flowchart illustrating the adaptive task allocation method in intelligent meeting collaboration;

[0055] Figure 2 A flowchart illustrating the process of establishing a collaborative task group aggregation and knowledge sharing mechanism. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0058] Figure 1 This is a flowchart illustrating the task adaptive allocation method in intelligent meeting collaboration according to an embodiment of the present invention.

[0059] Adaptive task allocation methods in intelligent meeting collaboration include:

[0060] Based on the knowledge coverage relationship between the task knowledge dependency characteristics of the meeting and the knowledge graph characteristics of the meeting participants, a knowledge matching network is constructed and the knowledge propagation path is deduced. The necessary knowledge nodes and knowledge gap nodes for task execution are identified, the knowledge complementarity index among participants is calculated, and the knowledge dependency analysis results are obtained.

[0061] Based on the knowledge dependency parsing results and task collaboration granularity characteristics, the task set is reconstructed in terms of collaboration granularity through a two-way adjustment mechanism of task decomposition and aggregation. When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks and a knowledge transfer interface between sub-tasks is established according to the dependency order of the required knowledge nodes.

[0062] When the overlap of the required knowledge nodes of multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, the multiple tasks are aggregated into a collaborative task group and a knowledge sharing mechanism is established to obtain the task structure after the collaborative granularity is reconstructed.

[0063] Based on the reconstructed task structure at the collaborative granularity and the cognitive switching cost characteristics of the participants, a cognitive state transition diagram is constructed. Based on the knowledge transfer interface, a task execution sequence with the minimum cognitive switching cost is planned for each participant in the cognitive state transition diagram. The task execution sequence is aligned with the time window based on the participants' historical task execution efficiency characteristics, and a task allocation instruction is generated.

[0064] In one optional implementation, based on the knowledge coverage relationship between the task knowledge dependency characteristics of the meeting and the knowledge graph characteristics of the meeting participants, a knowledge matching network is constructed and a knowledge propagation path is deduced. Essential knowledge nodes and knowledge gap nodes for task execution are identified, and the knowledge complementarity index among participants is calculated. The knowledge dependency resolution results include:

[0065] The knowledge items in the task knowledge dependency features are represented as task knowledge nodes, and the knowledge items in the knowledge graph features are represented as participant knowledge nodes. Based on the semantic similarity and knowledge hierarchy relationship between the task knowledge nodes and the participant knowledge nodes, the knowledge matching network containing a set of task knowledge nodes, a set of participant knowledge nodes, and cross-set association edges is constructed.

[0066] In the knowledge matching network, the task knowledge node is selected as the starting node. The reachable participant knowledge nodes are traversed by multi-hop propagation along the associated edges. The knowledge propagation path from the task knowledge node to the participant knowledge node is identified. The knowledge reachability metric of the task knowledge node is calculated based on the path length of the knowledge propagation path and the weight of the associated edges on the path.

[0067] Task knowledge nodes whose knowledge accessibility metric exceeds a preset accessibility threshold are marked as required knowledge nodes, and the remaining task knowledge nodes are marked as knowledge gap nodes;

[0068] Extract a subset of participant knowledge nodes covering the knowledge gap node from the knowledge matching network and obtain the knowledge association strength between participants from it. Calculate the knowledge complementarity index of participants on the knowledge gap node based on the knowledge association strength.

[0069] The knowledge gap nodes are combined with the knowledge complementarity index to form the knowledge dependency analysis result.

[0070] A unified node-based representation is applied to the knowledge items in the task knowledge dependency features and participant knowledge graph features. Each knowledge item involved in the task knowledge dependency features is abstracted into a task knowledge node, which carries information about the knowledge type, knowledge level, and dependencies required to complete the task. Simultaneously, knowledge items in each participant's knowledge graph features are abstracted into participant knowledge nodes, each containing attributes such as the participant's mastery level, knowledge depth, and knowledge association range in the corresponding knowledge domain. After completing the node-based representation, cross-set association edges are established between the two types of node sets based on the semantic similarity and knowledge hierarchy relationships between task knowledge nodes and participant knowledge nodes, thus forming a knowledge matching network. Semantic similarity is calculated by using a pre-trained semantic encoding model to vectorize the textual descriptions of knowledge items, and cosine similarity is used to measure the semantic closeness between two knowledge items. Knowledge hierarchy relationships are determined based on existing hierarchical relationships in the domain ontology or knowledge graph. When a hierarchical inclusion relationship exists between a task knowledge node and a participant knowledge node, or when the semantic similarity exceeds a set threshold, an association edge is established between them and assigned a corresponding edge weight. The resulting knowledge matching network includes a set of task knowledge nodes, a set of participant knowledge nodes, and cross-set association edges connecting the two types of nodes, fully depicting the coverage relationship between task knowledge requirements and participant knowledge reserves.

[0071] After the knowledge matching network is constructed, knowledge propagation path deduction is performed. Starting with the task knowledge node, a multi-hop propagation traversal is performed along the associated edges towards the participant knowledge nodes. Each hop extends along the outgoing edges of the current node to adjacent nodes, until all reachable participant knowledge nodes are traversed. This multi-hop propagation design allows knowledge to be indirectly covered in the network via intermediate nodes. That is, when a task knowledge node cannot directly establish an associated edge with a participant knowledge node, indirect propagation can be achieved through semantically similar intermediate knowledge nodes, thereby capturing implicit coverage relationships between knowledge. For each knowledge propagation path from the task knowledge node to the participant knowledge node, the knowledge reachability metric corresponding to that path is calculated based on the path length and the weights of the associated edges on the path. Specifically, the shorter the path length and the higher the weights of the associated edges on the path, the larger the corresponding knowledge reachability metric, indicating a stronger coverage ability of the participant for that task knowledge node. If multiple propagation paths exist from the same task knowledge node leading to different participant knowledge nodes, the maximum value of the knowledge reachability metric among all paths is taken as the final knowledge reachability metric for that task knowledge node, denoted as . ,in This is the index number for the task knowledge node.

[0072] Set a preset accessibility threshold ,Will The task knowledge node is marked as a required knowledge node, indicating that the knowledge node is fully covered by the knowledge graph of the participant group and can be effectively supported during task execution; Task knowledge nodes are marked as knowledge gap nodes, indicating that the current participant group's coverage of these knowledge nodes is insufficient, resulting in knowledge gaps. The classification of required knowledge nodes and knowledge gap nodes directly reflects the degree of matching between task knowledge requirements and participants' knowledge reserves, providing a basis for subsequent task decomposition and collaborative granularity restructuring. A preset reachability threshold is used. The value can be dynamically adjusted based on the importance of the meeting task and the strictness of the knowledge requirements. For task scenarios with high knowledge requirements, it can be appropriately increased. To ensure sufficient knowledge coverage.

[0073] For the identified knowledge gap nodes, a subset of participant knowledge nodes that can cover these gap nodes is extracted from the knowledge matching network. Coverage refers to the existence of a connection edge between a participant's knowledge node and the knowledge gap node, and the weight of this connection edge meets a certain lower bound requirement, meaning the participant has a certain level of relevant knowledge accumulation in the knowledge gap domain, although not yet reaching the reachability threshold for complete coverage. After extracting the subset of participant knowledge nodes covering the knowledge gap nodes, the knowledge association strength between different participants is obtained from this subset. Knowledge association strength characterizes the degree of complementarity between two participants' knowledge differences at the knowledge gap node, i.e., the synergistic potential between one participant's knowledge depth in a certain knowledge gap dimension and another participant's knowledge breadth in that dimension. Based on the knowledge association strength, a knowledge complementarity index of participants at the knowledge gap node is calculated, denoted as [Insert index here]. ,in and These are the index numbers of the two participants. The higher the value, the more participants... With participants The stronger the complementarity at knowledge gap nodes, the higher the efficiency of knowledge collaboration when the two parties jointly undertake tasks involving the knowledge gap. The calculation of the knowledge complementarity index comprehensively considers the participants' knowledge coverage, knowledge depth differences, and knowledge correlation strength at the knowledge gap nodes, and can fully reflect the collaborative potential between participants at the knowledge level.

[0074] The identified set of knowledge gap nodes is compared with the calculated knowledge complementarity index among participants. The combination of these elements constitutes the knowledge dependency resolution result. Within this result, the set of knowledge gap nodes clarifies the specific knowledge dimensions where knowledge blind spots exist in the current task execution, while the knowledge complementarity index quantifies the collaborative ability of different participant combinations to fill these knowledge gaps. This resolution result provides precise knowledge-level input for subsequent bidirectional adjustments to task decomposition and aggregation based on task collaboration granularity characteristics, ensuring that the task allocation scheme has a reliable theoretical basis at the knowledge coverage level. In actual meeting scenarios, the participants' knowledge graph features can be dynamically updated through various means such as historical meeting records, professional background information, and real-time knowledge assessment, thereby ensuring the timeliness and accuracy of the knowledge matching network and further improving the reliability of the knowledge dependency resolution result.

[0075] In one optional implementation, when the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks, and a knowledge transfer interface between the sub-tasks is established according to the dependency order of the required knowledge nodes, including:

[0076] When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the set of necessary knowledge nodes corresponding to the current task is extracted. A necessary knowledge node dependency graph is constructed based on the predecessor and successor relationships between knowledge nodes in the set of necessary knowledge nodes. Knowledge nodes without predecessor nodes are identified as dependency starting points and knowledge nodes without successor nodes are identified as dependency ending points in the necessary knowledge node dependency graph. The dependency order of necessary knowledge nodes is determined by traversing the path from the dependency starting point to the dependency ending point.

[0077] The set of necessary knowledge nodes is sequentially segmented according to the dependency order of the necessary knowledge nodes. The necessary knowledge nodes in each segment are combined with the knowledge gap nodes that cover the necessary knowledge nodes in the segment to form a subtask knowledge requirement unit. The current task is decomposed into multiple subtasks based on the subtask knowledge requirement unit.

[0078] For adjacent segments in the required knowledge node dependency graph that have predecessor-successor relationships, the mapping relationship between the required knowledge node identifiers output by the preceding subtask and the required knowledge node identifiers depended on by the subsequent subtask is recorded, and a knowledge transfer interface is established.

[0079] When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task needs to be structurally decomposed. The criterion for triggering the decomposition operation is: statistically analyzing the set of knowledge gap nodes corresponding to the current task. If its size exceeds the upper limit of knowledge nodes that any single participant can cover in the current meeting context, then the task is deemed impossible to complete independently by a single person, and the task decomposition process must be initiated. At this point, the essential knowledge node set corresponding to the current task is first extracted. This set contains all knowledge nodes indispensable for completing the current task, including nodes already covered by participants and nodes where gaps still exist.

[0080] After obtaining the set of necessary knowledge nodes, it is necessary to analyze the predecessor-successor relationships between the knowledge nodes within the set. A predecessor-successor relationship refers to the fact that in a knowledge system, the understanding or application of one knowledge node requires the mastery of another knowledge node; that is, the output of a predecessor node is the input dependency of a successor node. Based on this directed dependency relationship, all nodes in the set of necessary knowledge nodes and their predecessor-successor edges are constructed into a necessary knowledge node dependency graph. This graph is a directed acyclic graph structure, which can fully express the logical execution order between knowledge nodes. In the necessary knowledge node dependency graph, knowledge nodes with an in-degree of zero (i.e., no predecessor node) are identified as dependency starting points, and knowledge nodes with an out-degree of zero (i.e., no successor node) are identified as dependency ending points. The dependency starting point represents the beginning of the task knowledge chain, and the dependency ending point represents the end of the knowledge chain.

[0081] By performing a path traversal from the dependency start point to the dependency end point on the dependency graph of the required knowledge nodes, the dependency order of all required knowledge nodes can be determined. The path traversal process uses topological sorting to ensure that each knowledge node is included in the sorted sequence only after all its predecessor nodes have been processed. Let the ordered sequence of required knowledge nodes obtained after topological sorting be... ,in The total number of required knowledge nodes. Indicates the first in the sequence There are several knowledge nodes in the sequence, and the nodes that appear earlier in the sequence are more likely to be dependent on the nodes that appear later in the sequence.

[0082] After obtaining the dependency order, the set of necessary knowledge nodes is sequentially segmented according to this order. The core principle of segmentation is: knowledge nodes within the same segment should be in close proximity at the knowledge dependency level and can be processed collectively by participants with corresponding knowledge backgrounds; there should be clear knowledge dependency boundaries between adjacent segments, that is, the output knowledge of the previous segment is the input dependency of the next segment. Segmentation can be implemented based on key dependency cut points in the dependency graph—when there is a node in the dependency graph whose predecessor and successor nodes have a large knowledge domain span, or whose required participant knowledge background differs significantly from that of the predecessor node, this node can be used as a segmentation boundary. Assume the segmentation result will be an ordered sequence. Divided into The segment, the first Each segment is denoted as ,but constituting a pair They completely cover each other and do not overlap.

[0083] For each segment The necessary knowledge nodes within each segment are combined with the knowledge gap nodes that can cover these nodes to form a subtask knowledge requirement unit. Each subtask knowledge requirement unit clearly describes the input requirements and outputs of the subtask in terms of knowledge dimension, making the knowledge boundaries of each subtask clearly definable. Based on the subtask knowledge requirement units corresponding to each segment, the current task is decomposed into... Each subtask corresponds to a segment, and each subtask is responsible for the knowledge processing within that segment. The execution order of the subtasks is consistent with the dependency order of the segments, i.e., the subtasks... The execution must be performed in the subtask You can only begin after completing and outputting the corresponding knowledge nodes (for adjacent subtasks that have direct dependencies).

[0084] After the subtasks are decomposed, a knowledge transfer interface needs to be established between adjacent segments that have predecessor-successor relationships. The knowledge transfer interface is established as follows: for segments in the dependency graph of necessary knowledge nodes... With segmentation There are cross-segment dependency edges between them, recording the segments. The required knowledge node identifiers output by the corresponding subtask (i.e., the preceding subtask), and the segmentation. The necessary knowledge node identifiers that the corresponding subtasks (i.e., subsequent subtasks) depend on are used to establish a mapping relationship between them. This mapping relationship expresses which knowledge nodes the preceding subtask needs to pass to the subsequent subtask, and which knowledge nodes the subsequent subtask needs to receive from the preceding subtask, thus forming a complete knowledge transfer interface.

[0085] The knowledge transfer interface serves as a conduit for knowledge connection and state transfer during actual execution. After a preceding subtask is completed, its resulting knowledge nodes are written to a shared knowledge middleware layer via the knowledge transfer interface. Subsequent subtasks read the required preceding knowledge nodes from this middleware layer before starting, ensuring the continuity and integrity of the knowledge chain. This mapping-based interface design explicitly manages the knowledge dependencies between subtasks, preventing subsequent subtasks from failing due to incomplete knowledge transfer. Simultaneously, the knowledge transfer interface provides a basis for the allocation of subsequent participants—participants responsible for preceding subtasks must output their generated knowledge node identifiers and corresponding content according to the interface specification after completing their tasks, while participants responsible for subsequent subtasks receive and utilize these knowledge nodes according to the interface specification, thus achieving knowledge relay collaboration among participants.

[0086] Throughout the task decomposition process, the dependency graph of essential knowledge nodes serves as the core data structure, permeating the entire process from determining the dependency order to segmentation and establishing knowledge transfer interfaces. The accuracy of the dependency graph directly determines the rationality of subtask decomposition. Therefore, when constructing the dependency graph, it is necessary to ensure the completeness and accuracy of predecessor-successor relationships, avoiding the omission of key dependency edges or the introduction of incorrect dependency directions. For cases where there are multiple parallel paths in the dependency graph, the segmentation operation must comprehensively consider the distribution of knowledge nodes on each path, ensuring that the segmentation boundaries can reasonably divide all parallel paths, and establishing corresponding knowledge transfer interfaces for cross-segment dependency edges on each path to guarantee the complete transfer of parallel knowledge flow.

[0087] In one optional implementation, when the overlap of the required knowledge nodes of multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, the multiple tasks are aggregated into a collaborative task group and a knowledge sharing mechanism is established, including:

[0088] Select any two tasks from the set of tasks to be assigned as a task pair, extract the sets of necessary knowledge nodes corresponding to the two tasks respectively, and calculate the overlap of necessary knowledge nodes of the task pair based on the ratio of the number of knowledge nodes in the intersection of the two sets of necessary knowledge nodes to the number of knowledge nodes in the union.

[0089] Traverse all task pairs in the set of tasks to be assigned, and filter out task pairs whose required knowledge node overlap exceeds a preset overlap threshold as candidate aggregation task pairs;

[0090] Extract the knowledge gap node sets corresponding to the two tasks in the candidate aggregation task pair, identify the previous participant subset that can cover the knowledge gap node set of the previous task and the next participant subset that can cover the knowledge gap node set of the next task from the participant set, and calculate the knowledge complementarity index between the previous participant subset and the next participant subset.

[0091] When the knowledge complementarity index exceeds the preset complementarity threshold, the two tasks in the candidate aggregation task pair are aggregated into a collaborative task group.

[0092] Extract the intersection knowledge nodes corresponding to the collaborative task groups as shared knowledge nodes, establish an access mapping relationship between the shared knowledge nodes and the participant knowledge nodes, and establish a knowledge sharing mechanism based on the access mapping relationship.

[0093] like Figure 2 As shown, the method includes:

[0094] Select any two tasks from the set of tasks to be assigned. and Form a task pair and extract the necessary knowledge node sets corresponding to these two tasks respectively. and Required knowledge node overlap It is calculated based on the ratio of the number of nodes in the intersection of two sets to the number of nodes in the union, specifically expressed as:

[0095]

[0096] in, This represents the number of nodes in the intersection of two sets of required knowledge nodes. This represents the number of nodes in the union of two sets of required knowledge nodes. The closer this ratio is to 1, the higher the overlap in knowledge requirements between the two tasks, indicating their potential to be aggregated into a collaborative task group. This calculation method is essentially a measure of the similarity between two sets, effectively reflecting the degree of commonality in knowledge requirements between tasks and avoiding participants repeatedly investing the same knowledge preparation work due to highly repetitive knowledge requirements.

[0097] Iterate through all possible task pairs in the set of tasks to be assigned, calculate the overlap of required knowledge nodes for each task pair, and compare it with a preset overlap threshold. Comparison. When When the task pair is marked as a candidate aggregation task pair, it is added to the candidate aggregation task pair set. Medium. Preset overlap threshold. The value should be determined comprehensively based on the knowledge intensity of the meeting tasks and the knowledge coverage of the participants. It can usually be configured in the range of 0.5 to 0.8 to ensure that the aggregated collaborative task groups have substantial commonalities in knowledge requirements, rather than meaningless task aggregation due to the threshold being too low.

[0098] for Each candidate aggregation task in the pair Further extract the knowledge gap node sets corresponding to each of the two tasks. and Knowledge gap nodes refer to essential knowledge nodes identified during the knowledge matching network construction phase that cannot be directly covered by the current participants' knowledge graphs. (From the participant set) Identify the coverage The subset of participants that represents all or most of the knowledge gap nodes is denoted as the previous participant subset. Similarly, identify those that can cover The subset of participants that represents all or most of the knowledge gap nodes is denoted as the next participant subset. The term "coverage" refers to the existence of a knowledge node in the participant's knowledge graph that corresponds to the target knowledge gap node, and the knowledge accessibility metric between the two meeting a preset accessibility threshold. Requirements.

[0099] In obtaining and Next, the knowledge complementarity index between the two participant subsets is calculated. The calculation method of this indicator is similar to the knowledge complementarity indicator among participants. The computational logic remains consistent, that is, it comprehensively considers the degree to which the two participant subsets mutually compensate for each other at the knowledge gap nodes, specifically reflected in... Chinese participants coverage and Chinese participants The comprehensive assessment results of coverage capability. When Exceeding the preset complementary threshold This indicates that the two participant subsets have a substantial complementary relationship in terms of knowledge and capabilities, thus linking the candidate aggregation task to... The two tasks in the process were formally merged into a collaborative task group. and will and The union of these sets is determined as the set of candidate participants for the collaborative task group.

[0100] Collaborative task group Once formed, a corresponding knowledge-sharing mechanism needs to be established to support the flow of knowledge during collaborative execution. (Extraction) The corresponding set of intersecting knowledge nodes As a collection of shared knowledge nodes Shared knowledge nodes are the knowledge foundation upon which both tasks depend. During collaborative execution, they do not require repeated preparation by different participants; instead, they are accessed through a unified access mapping relationship.

[0101] When establishing the access mapping relationship between shared knowledge nodes and participant knowledge nodes, for Each shared knowledge node in Iterate through each participant in the set of candidate participants for collaborative task groups, and retrieve their knowledge graph from the relevant information. Corresponding knowledge nodes and will Write it as an access mapping record in the access mapping table The access mapping table structure needs to support many-to-many mapping relationships, meaning that the same shared knowledge node can be mapped to the knowledge nodes of multiple participants, and the knowledge node of the same participant can also be mapped to multiple shared knowledge nodes. This design ensures that during collaborative execution, when a participant needs to access a shared knowledge node, the access mapping table can quickly locate the participant who possesses that knowledge, thereby achieving on-demand knowledge flow and efficient sharing.

[0102] Based on access mapping table When establishing a knowledge-sharing mechanism, the priority order of knowledge sharing is further determined for each pair of participants in the collaborative task group. The priority is determined based on the knowledge accessibility metric between the participant's knowledge node and the shared knowledge node. Participants with higher accessibility metrics have higher knowledge-sharing priority on that shared knowledge node and are given priority in outputting and transmitting relevant content during collaborative execution. In this way, the knowledge-sharing mechanism not only clarifies the access path to shared knowledge but also standardizes the knowledge transfer order and responsibility allocation within the collaborative task group, thereby effectively reducing collaborative friction caused by redundant knowledge preparation or knowledge access conflicts during execution and improving overall task execution efficiency.

[0103] After completing the aggregation judgment and knowledge sharing mechanism for all candidate aggregation task pairs, all formed collaborative task groups and independent tasks that have not triggered aggregation conditions are incorporated into the task structure after collaborative granularity reconstruction, providing structured input for the subsequent construction of cognitive state transition diagrams and planning of task execution sequences.

[0104] In one optional implementation, constructing a cognitive state transition graph based on the task structure reconstructed at the collaborative granularity and the cognitive switching cost characteristics of the participants includes:

[0105] Extract all subtask nodes and collaborative task group nodes from the task structure after the collaborative granularity reconstruction, and use the subtask nodes and collaborative task group nodes as cognitive state nodes.

[0106] For each cognitive state node, extract the set of required knowledge nodes and the set of knowledge gap nodes corresponding to the cognitive state node;

[0107] For all cognitive state nodes, calculate the knowledge difference between the set of necessary knowledge nodes corresponding to the previous cognitive state node and the set of necessary knowledge nodes corresponding to the next cognitive state node, and identify a subset of participants from the participant set that can simultaneously cover the set of knowledge gap nodes of the previous cognitive state node and the set of knowledge gap nodes of the next cognitive state node.

[0108] For each participant in the subset of participants, the cognitive switching cost feature of the participant is extracted, and the cognitive switching cost from the previous cognitive state node to the next cognitive state node is calculated based on the knowledge difference and the cognitive switching cost feature.

[0109] A directed edge is established between the previous cognitive state node and the next cognitive state node. The cognitive switching cost is used as the edge weight of the directed edge. A cognitive state transition graph is constructed based on all cognitive state nodes and the directed edge.

[0110] After refactoring the collaboration granularity, a task structure containing sub-task nodes and collaborative task group nodes was obtained. The first step in constructing the cognitive state transition diagram is to extract all node types from this task structure. Specifically, each sub-task node generated by the decomposition operation, and each collaborative task group node formed by the aggregation operation, are included in the set of cognitive state nodes. Each cognitive state node represents a cognitive working state of the participant during task execution, and the switching between different nodes corresponds to the participant's shift in knowledge focus, attentional resources, and thinking framework. Unifying all sub-task nodes and collaborative task group nodes into cognitive state nodes is the foundation for subsequently establishing state transition relationships.

[0111] For each cognitive state node, two types of knowledge sets need to be extracted: a set of necessary knowledge nodes and a set of knowledge gap nodes. The set of necessary knowledge nodes, derived from the results of the preceding knowledge dependency resolution stage, describes the knowledge elements necessary to complete the task corresponding to this cognitive state node. The set of knowledge gap nodes describes knowledge nodes in the current participant group that are not yet fully covered. These two sets together characterize the knowledge requirement profile of each cognitive state node, providing input for subsequent calculations of knowledge differences between states.

[0112] Among all cognitive state node pairs, it is necessary to calculate the knowledge difference between adjacent states. Let the previous cognitive state node be... The next cognitive state node is , The corresponding set of required knowledge nodes is denoted as , The corresponding set of required knowledge nodes is denoted as The knowledge difference between the two Defined as the ratio of the size of the symmetric difference of two sets to the size of the size of the union of the two sets, i.e. ,in This indicates the number of elements in the set. This metric reflects the state... Switch to state The degree to which the focus of knowledge shifts at that time. When the value is close to 0, it indicates that the knowledge requirements of the two states highly overlap, and the switching cost is low; when... When the value is close to 1, it indicates that the knowledge requirements of the two states are almost completely different, and the switching cost is high.

[0113] Simultaneously, in calculating the knowledge dissimilarity, it is also necessary to identify a subset of participants from the participant set that simultaneously covers both the knowledge gap node set of the previous cognitive state node and the knowledge gap node set of the next cognitive state node. Let the previous cognitive state node be... The set of knowledge gap nodes is The next cognitive state node The set of knowledge gap nodes is Then the subset of participants that meet the conditions This refers to its knowledge graph being able to simultaneously cover and The set of participants for all nodes in the process. The purpose of identifying this subset of participants is that the calculation of the cognitive switching cost needs to be based on specific participants. Only participants who can take on two adjacent states are qualified to perform tasks on the directed edge.

[0114] For a subset of participants For each participant, their cognitive switching cost characteristics are extracted. These characteristics are derived from the participants' historical task execution records and include multi-dimensional quantitative indicators such as the time delay, the increase in error rate, and the preparation time required for attention refocusing when switching between different knowledge domains. Let the participants... The cognitive switching cost feature vector is This includes a switching time cost component. Error rate cost component And preparation time and cost Based on knowledge differences Features of cognitive switching costs Calculate participants From cognitive state nodes Switch to The cost of cognitive switching Its calculation method is to fuse a weighted combination of knowledge difference and participant switching cost characteristics, that is... ,in , , These are weighting coefficients for time cost, error rate cost, and preparation time cost, respectively, with a sum of 1. The values ​​of these weighting coefficients can be configured according to the different emphases on timeliness, accuracy, and consistency of the meeting task; by default, all three are equally distributed.

[0115] When participants subset When there are multiple participants, the minimum cognitive switching cost of all participants is taken as the edge weight of the directed edge, i.e. The logic behind this design is that, during the task allocation phase, priority should be given to having the participant with the lowest cognitive switching cost take over the transition between adjacent states, in order to minimize the cognitive burden during the overall collaboration process. If the participant subset... If the set is empty, meaning there are no participants who can simultaneously cover the knowledge gaps in both states, then the weight of the directed edge is set to positive infinity. This indicates that the state transition path cannot be directly implemented under the current participant configuration, and an additional knowledge supplementation mechanism or a readjustment of the collaboration granularity is required.

[0116] After calculating the knowledge difference and cognitive switching cost for all adjacent cognitive state node pairs, in the previous cognitive state node... With the next cognitive state node Establish directed edges between them Set the corresponding edge weights Assign a directed edge. The directionality of the directed edge reflects the temporal constraint of task execution, meaning that participants must complete the tasks corresponding to each cognitive state node in the order prescribed by the knowledge dependencies. Integrate all cognitive state nodes with all directed edges to form a complete cognitive state transition graph. This graph uses cognitive state nodes as the vertex set and directed edges and their weights as the edge set, fully describing the cognitive state transition space of participants throughout the entire meeting task execution process.

[0117] Cognitive state transition diagram The construction of this model provides a structured search space for subsequent task execution sequence planning. In this diagram, planning the task execution sequence with the minimum cognitive switching cost for each participant is equivalent to... This paper proposes a shortest path algorithm to find the shortest path from the initial cognitive state node to the final cognitive state node. The path length is obtained by summing the edge weights of the directed edges along the path. By executing the shortest path algorithm on the cognitive state transition graph, a task execution sequence with the minimum cognitive burden can be obtained for each participant. This reduces cognitive fatigue during the meeting collaboration process while ensuring the quality of task completion, thereby improving the overall collaboration efficiency.

[0118] In one optional implementation, planning the task execution sequence with minimum cognitive switching cost for each participant in the cognitive state transition graph based on the knowledge transfer interface includes:

[0119] For each participant, extract all cognitive state nodes that the participant can cover the set of knowledge gap nodes from the cognitive state transition graph as the reachable cognitive state node set;

[0120] Select an initial cognitive state node from the set of reachable cognitive state nodes. Starting from the initial cognitive state node, traverse the directed edges connecting the nodes in the set of reachable cognitive state nodes in the cognitive state transition graph and extract the edge weights. Accumulate the edge weights of all directed edges on the path from the initial cognitive state node to the target cognitive state node to obtain the total path cost.

[0121] For all cognitive state nodes in the reachable cognitive state node set, the path with the minimum total cost covering all cognitive state nodes in the reachable cognitive state node set is calculated as a candidate task execution sequence. Based on the mapping relationship between knowledge transfer source nodes and knowledge transfer target nodes recorded in the knowledge transfer interface, the cognitive switching cost between adjacent cognitive state nodes in the candidate task execution sequence is adjusted. The candidate task execution sequence with the minimum adjusted total cost is determined as the task execution sequence of the participant.

[0122] For each participant, from the cognitive state transition diagram Extract all cognitive state nodes that the participant can cover their set of knowledge gap nodes, and form a set of reachable cognitive state nodes. Specifically, for participants traversal For all cognitive state nodes in the graph, determine whether the set of required knowledge nodes corresponding to each node satisfies a coverage relationship with the participant's knowledge graph features. That is, whether the knowledge scope covered by the participant's knowledge graph includes all the required knowledge nodes for that cognitive state node. If the coverage relationship is satisfied, then include that cognitive state node in the graph. This process ensures that subsequent path planning is carried out only within the cognitive capacity of the participants, avoiding the inclusion of tasks beyond their knowledge and ability boundaries in their execution sequence, thereby guaranteeing the feasibility of task allocation.

[0123] After obtaining the set of reachable cognitive state nodes Afterwards, from Selecting the initial cognitive state node The selection principle for initial cognitive state nodes is: prioritize nodes that are closest to the participant's current knowledge state and have the lowest switching cost; that is, start from the participant's knowledge graph characteristics and calculate... The cognitive switching cost of each node relative to the participant's current knowledge state is calculated, and the node with the lowest cost is selected as the [node name / node name]. .by Starting from, in Middle traversal connection Directed edges of each node Extract the edge weight corresponding to each directed edge. Total path cost It is obtained by accumulating the edge weights of all directed edges traversed along the path, i.e. The summation range is from the initial cognitive state node. All directed edges traversed to reach the target cognitive state node.

[0124] against All cognitive state nodes need to find a path that covers... All nodes and total path cost The shortest path is selected as the candidate task execution sequence. This problem is formally equivalent to finding the shortest Hamiltonian path covering a given set of nodes in a directed weighted graph. Since... The size of the problem is usually limited by the capabilities of the participants, and can be solved precisely using dynamic programming. Specifically, let the state be... The set of visited nodes is The current node is The minimum total path cost at time, where , The state transition equation is: ,in This indicates removing the current node from the visited set. The subset after. When cover When considering all nodes, the minimum value is taken for all possible termination nodes, thus obtaining the optimal candidate task execution sequence that covers all reachable cognitive state nodes. .

[0125] After obtaining the candidate task execution sequence Then, based on the mapping relationship between the knowledge transfer source node and the knowledge transfer target node recorded in the knowledge transfer interface, further... The cognitive switching cost between adjacent cognitive state nodes is adjusted. The knowledge transfer interface, established during the task decomposition phase according to the dependency order of necessary knowledge nodes, records the flow of knowledge among participants, i.e., which knowledge nodes are transferred from participants executing preceding tasks to participants executing subsequent tasks. If... Two adjacent cognitive state nodes and There is a knowledge transfer interface between them, indicating that from Switch to At times, some knowledge gaps can be filled by previous participants through an interface, thereby reducing the burden on participants. The amount of knowledge that needs to be mastered independently during this switching process. In this case, the corresponding cognitive switching cost. The reduction should be based on the actual difference in knowledge required, with the specific reduction amount depending on the number of knowledge nodes covered by the knowledge transfer interface. The ratio determines the cost reduction; that is, the higher the proportion of nodes covered by the knowledge transfer interface, the greater the reduction in switching costs. After adjustment, recalculate. Total path cost The adjusted candidate task execution sequence cost is obtained.

[0126] If multiple candidate task execution sequences are generated during the initial path planning phase (e.g., multiple paths with the same or similar costs exist during the dynamic programming backtracking phase), then after adjusting the knowledge transfer interface, the total path cost after adjusting all candidate sequences will be... Compare, The smallest candidate task execution sequence was determined as the participant. Final task execution sequence If multiple candidate sequences with the same adjusted cost exist, the sequence with higher historical execution efficiency is selected as the final result, based on the participants' historical task execution efficiency characteristics as a secondary ranking criterion. This ensures that, given comparable cognitive switching costs, the sequence with better execution efficiency is prioritized.

[0127] right Each participant independently executes the above planning process, resulting in a task execution sequence for each participant. In the subsequent time window alignment phase, the task execution sequences of each participant will be further coordinated by incorporating historical task execution efficiency characteristics. This ensures effective continuity of task execution across multiple participants in the time dimension, preventing excessive waiting costs at the knowledge transfer interface due to delays in the task sequence of a single participant. The entire planning process embeds the mapping relationship of the knowledge transfer interface into a path cost adjustment mechanism, ensuring that the planning of the task execution sequence not only considers the cognitive switching burden of each participant but also fully leverages the knowledge collaboration relationships among participants, achieving a comprehensive optimization of cognitive switching costs and collaboration efficiency.

[0128] In one optional implementation, the task execution sequence is aligned with a time window based on the participants' historical task execution efficiency characteristics to generate task allocation instructions, including:

[0129] Extract the set of necessary knowledge nodes and task complexity features corresponding to each cognitive state node from the task execution sequence, and predict the expected execution time of the cognitive state node based on the set of necessary knowledge nodes, the task complexity features and the historical task execution efficiency features;

[0130] The cognitive switching cost between adjacent cognitive state nodes in the task execution sequence is converted into a time buffer. The start time of the available time window is used as the time starting point, and execution time periods are allocated to the cognitive state nodes in the task execution sequence in sequence within the duration of the available time window.

[0131] When the execution time of a cognitive state node in the task execution sequence exceeds the duration of the available time window, a subset of cognitive state nodes that can be delayed in execution is identified from the task execution sequence, and the subset of cognitive state nodes that can be delayed in execution is mapped to the next available time window of the participant.

[0132] Based on the task execution sequence, the time period corresponding to each cognitive state node and the time window mapping relationship between the subset of cognitive state nodes that can be delayed, task allocation instructions are generated for the meeting participants.

[0133] The necessary knowledge node set corresponding to each cognitive state node is extracted from the task execution sequence, and the task complexity feature of that node is obtained simultaneously. The task complexity feature encompasses the number of knowledge nodes involved in the task, the depth of the dependency hierarchy between knowledge nodes, and the degree of structure of the task itself. Historical task execution efficiency features are derived from the actual completion time statistics of participants for similar knowledge domains and similar complexity levels in past meetings or collaboration records. The size of the necessary knowledge node set, the quantified value of the complexity feature, and the historical efficiency feature are jointly input into the prediction model to obtain the expected execution time of that cognitive state node. ,in This serves as the index for the current cognitive state node. During the prediction of expected execution time, historical efficiency features are used as individualized correction factors, ensuring that the expected execution time for the same task varies among different participants, thus avoiding execution bias caused by a "one-size-fits-all" time allocation approach.

[0134] The cost of cognitive switching between adjacent cognitive state nodes The calculations have already been completed during the path planning phase; here, they are converted into time buffers. The conversion method is based on the three components of cognitive switching costs: switching time cost, error rate cost, and preparation time cost. These are converted into time units according to their respective physical meanings. The switching time cost and preparation time cost components already have a time dimension and can be directly added together. The error rate cost component is converted using the average historical error recovery time to obtain its corresponding time equivalent. The sum of these three components yields the time buffer. The introduction of a time buffer ensures that there is a sufficient transition interval between adjacent tasks, preventing a decline in execution quality due to insufficient cognitive state switching.

[0135] Starting time of the available time window As the starting point of time, the duration of the available time window is denoted as... This represents the total time available for participants to execute tasks during the current meeting. Execution time slots are allocated sequentially to each cognitive state node according to their order in the task execution sequence. The execution time slot for the first cognitive state node begins at [time value missing]. The end time is The starting time of the second cognitive state node is The end time is This process continues, forming a time-segmentation scheme that covers the entire task execution sequence. The execution time of each cognitive state node is determined by its expected execution duration and the preceding time buffer, and time segments are not allowed to overlap.

[0136] When there is a cognitive state node in the task execution sequence, and the end time of its allocated execution time period exceeds [a certain threshold], This indicates that the currently available time window cannot accommodate all tasks in the sequence. In this case, it is necessary to identify a subset of cognitive state nodes that can be delayed from the task execution sequence. The criteria for determining deferred execution include the following dimensions: whether the cognitive state node is a prerequisite for other nodes in the knowledge transfer interface; if subsequent nodes depend on its output, it cannot be deferred; whether the task corresponding to the node is a critical path node in the collaborative task group; if so, it cannot be deferred; and whether the task priority label of the node is high priority; if so, it cannot be deferred. After satisfying the above constraints, the remaining cognitive state nodes are included in the candidate set of deferred execution nodes, and then the final subset of deferred execution cognitive state nodes is selected based on the principle of minimizing the impact on the overall task completion progress. .

[0137] Will Each cognitive state node in the algorithm is mapped to the participant's next available time window. The start time of the next available time window is denoted as . This value is derived from the participants' schedules or meeting scheduling information. The mapping process also needs to... Within the corresponding time window duration, the time period allocation logic is re-executed to ensure that the execution time periods of delayed nodes are legally arranged and do not exceed the boundaries within the new time window. If the next available time window still cannot accommodate all delayed nodes, the above mapping process is recursively executed until all cognitive state nodes obtain legal execution time periods.

[0138] After all cognitive state nodes have completed their execution time slot allocation, task allocation instructions for meeting participants are generated based on the task execution sequence, the execution time slot corresponding to each cognitive state node, and the time window mapping relationship of the subset of cognitive state nodes that can be delayed. The task allocation instructions include the participant identifier, task identifier, start and end times of the execution time slot, the time window number, and the knowledge sharing access permissions required for task execution (for participants in collaborative task groups, access mapping information for the shared knowledge node set must be added). For adjacent subtasks with knowledge transfer interfaces, the task allocation instructions must also indicate the triggering conditions for knowledge transfer, i.e., the timing and method of sending knowledge transfer notifications to subsequent subtask participants after the completion of the preceding subtask. The final generated task allocation instructions are stored in a structured format, supporting direct parsing and push to the corresponding participant's task panel by the meeting collaboration platform, achieving a closed-loop process from knowledge dependency analysis to time scheduling.

[0139] A second aspect of the present invention provides a task adaptive allocation system for intelligent meeting collaboration, comprising:

[0140] The dependency resolution unit is used to construct a knowledge matching network and deduce the knowledge propagation path based on the knowledge coverage relationship between the task knowledge dependency features of the meeting and the knowledge graph features of the meeting participants. It identifies the necessary knowledge nodes and knowledge gap nodes for task execution, calculates the knowledge complementarity index among participants, and obtains the knowledge dependency resolution results.

[0141] The granularity reconstruction unit is used to reconstruct the collaborative granularity of the task set based on the knowledge dependency parsing results and task collaboration granularity characteristics through a two-way adjustment mechanism of task decomposition and aggregation. When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks and a knowledge transfer interface between sub-tasks is established according to the dependency order of the necessary knowledge nodes.

[0142] The task aggregation unit is used to aggregate multiple tasks into a collaborative task group and establish a knowledge sharing mechanism when the overlap of the required knowledge nodes of multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, thereby obtaining the task structure after the collaborative granularity reconstruction.

[0143] The task allocation unit is used to construct a cognitive state transition diagram based on the task structure reconstructed by the collaborative granularity and the cognitive switching cost characteristics of the participants, plan a task execution sequence with minimum cognitive switching cost for each participant in the cognitive state transition diagram based on the knowledge transfer interface, align the task execution sequence with a time window based on the historical task execution efficiency characteristics of the participants, and generate a task allocation instruction.

[0144] A third aspect of the present invention provides an electronic device, comprising:

[0145] processor;

[0146] Memory used to store processor-executable instructions;

[0147] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0148] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0149] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A task adaptive allocation method in intelligent meeting collaboration, characterized in that, include: Based on the knowledge coverage relationship between the task knowledge dependency characteristics of the meeting and the knowledge graph characteristics of the meeting participants, a knowledge matching network is constructed and the knowledge propagation path is deduced. The necessary knowledge nodes and knowledge gap nodes for task execution are identified, the knowledge complementarity index among participants is calculated, and the knowledge dependency analysis results are obtained. Based on the knowledge dependency parsing results and task collaboration granularity characteristics, the task set is reconstructed in terms of collaboration granularity through a two-way adjustment mechanism of task decomposition and aggregation. When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks and a knowledge transfer interface between sub-tasks is established according to the dependency order of the required knowledge nodes. When the overlap of the required knowledge nodes of multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, the multiple tasks are aggregated into a collaborative task group and a knowledge sharing mechanism is established to obtain the task structure after the collaborative granularity is reconstructed. Based on the reconstructed task structure at the collaborative granularity and the cognitive switching cost characteristics of the participants, a cognitive state transition diagram is constructed. Based on the knowledge transfer interface, a task execution sequence with the minimum cognitive switching cost is planned for each participant in the cognitive state transition diagram. The task execution sequence is aligned with the time window based on the participants' historical task execution efficiency characteristics, and a task allocation instruction is generated.

2. The method according to claim 1, characterized in that, Based on the knowledge coverage relationship between the task knowledge dependency characteristics of the meeting and the knowledge graph characteristics of the participants, a knowledge matching network is constructed and the knowledge propagation path is deduced. Essential knowledge nodes and knowledge gap nodes for task execution are identified, and the knowledge complementarity index among participants is calculated. The knowledge dependency resolution results include: The knowledge items in the task knowledge dependency features are represented as task knowledge nodes, and the knowledge items in the knowledge graph features are represented as participant knowledge nodes. Based on the semantic similarity and knowledge hierarchy relationship between the task knowledge nodes and the participant knowledge nodes, the knowledge matching network containing a set of task knowledge nodes, a set of participant knowledge nodes, and cross-set association edges is constructed. In the knowledge matching network, the task knowledge node is selected as the starting node. The reachable participant knowledge nodes are traversed by multi-hop propagation along the associated edges. The knowledge propagation path from the task knowledge node to the participant knowledge node is identified. The knowledge reachability metric of the task knowledge node is calculated based on the path length of the knowledge propagation path and the weight of the associated edges on the path. Task knowledge nodes whose knowledge accessibility metric exceeds a preset accessibility threshold are marked as required knowledge nodes, and the remaining task knowledge nodes are marked as knowledge gap nodes; Extract a subset of participant knowledge nodes covering the knowledge gap node from the knowledge matching network and obtain the knowledge association strength between participants from it. Calculate the knowledge complementarity index of participants on the knowledge gap node based on the knowledge association strength. The knowledge gap nodes are combined with the knowledge complementarity index to form the knowledge dependency analysis result.

3. The method according to claim 1, characterized in that, When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks, and a knowledge transfer interface is established between the sub-tasks according to the dependency order of the required knowledge nodes, including: When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the set of necessary knowledge nodes corresponding to the current task is extracted. A necessary knowledge node dependency graph is constructed based on the predecessor and successor relationships between knowledge nodes in the set of necessary knowledge nodes. Knowledge nodes without predecessor nodes are identified as dependency starting points and knowledge nodes without successor nodes are identified as dependency ending points in the necessary knowledge node dependency graph. The dependency order of necessary knowledge nodes is determined by traversing the path from the dependency starting point to the dependency ending point. The set of necessary knowledge nodes is sequentially segmented according to the dependency order of the necessary knowledge nodes. The necessary knowledge nodes in each segment are combined with the knowledge gap nodes that cover the necessary knowledge nodes in the segment to form a subtask knowledge requirement unit. The current task is decomposed into multiple subtasks based on the subtask knowledge requirement unit. For adjacent segments in the required knowledge node dependency graph that have predecessor-successor relationships, the mapping relationship between the required knowledge node identifiers output by the preceding subtask and the required knowledge node identifiers depended on by the subsequent subtask is recorded, and a knowledge transfer interface is established.

4. The method according to claim 1, characterized in that, When the overlap of required knowledge nodes in multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, the multiple tasks are aggregated into a collaborative task group and a knowledge-sharing mechanism is established, including: Select any two tasks from the set of tasks to be assigned as a task pair, extract the sets of necessary knowledge nodes corresponding to the two tasks respectively, and calculate the overlap of necessary knowledge nodes of the task pair based on the ratio of the number of knowledge nodes in the intersection of the two sets of necessary knowledge nodes to the number of knowledge nodes in the union. Traverse all task pairs in the set of tasks to be assigned, and filter out task pairs whose required knowledge node overlap exceeds a preset overlap threshold as candidate aggregation task pairs; Extract the knowledge gap node sets corresponding to the two tasks in the candidate aggregation task pair, identify the previous participant subset that can cover the knowledge gap node set of the previous task and the next participant subset that can cover the knowledge gap node set of the next task from the participant set, and calculate the knowledge complementarity index between the previous participant subset and the next participant subset. When the knowledge complementarity index exceeds the preset complementarity threshold, the two tasks in the candidate aggregation task pair are aggregated into a collaborative task group. Extract the intersection knowledge nodes corresponding to the collaborative task groups as shared knowledge nodes, establish an access mapping relationship between the shared knowledge nodes and the participant knowledge nodes, and establish a knowledge sharing mechanism based on the access mapping relationship.

5. The method according to claim 1, characterized in that, Based on the task structure reconstructed at the collaborative granularity and the cognitive switching cost characteristics of the participants, a cognitive state transition diagram is constructed, including: Extract all subtask nodes and collaborative task group nodes from the task structure after the collaborative granularity reconstruction, and use the subtask nodes and collaborative task group nodes as cognitive state nodes. For each cognitive state node, extract the set of required knowledge nodes and the set of knowledge gap nodes corresponding to the cognitive state node; For all cognitive state nodes, calculate the knowledge difference between the set of necessary knowledge nodes corresponding to the previous cognitive state node and the set of necessary knowledge nodes corresponding to the next cognitive state node, and identify a subset of participants from the participant set that can simultaneously cover the set of knowledge gap nodes of the previous cognitive state node and the set of knowledge gap nodes of the next cognitive state node. For each participant in the subset of participants, the cognitive switching cost feature of the participant is extracted, and the cognitive switching cost from the previous cognitive state node to the next cognitive state node is calculated based on the knowledge difference and the cognitive switching cost feature. A directed edge is established between the previous cognitive state node and the next cognitive state node. The cognitive switching cost is used as the edge weight of the directed edge. A cognitive state transition graph is constructed based on all cognitive state nodes and the directed edge.

6. The method according to claim 1, characterized in that, Based on the knowledge transfer interface, the task execution sequence with minimum cognitive switching cost planned for each participant in the cognitive state transition diagram includes: For each participant, extract all cognitive state nodes that the participant can cover the set of knowledge gap nodes from the cognitive state transition graph as the reachable cognitive state node set; Select an initial cognitive state node from the set of reachable cognitive state nodes. Starting from the initial cognitive state node, traverse the directed edges connecting the nodes in the set of reachable cognitive state nodes in the cognitive state transition graph and extract the edge weights. Accumulate the edge weights of all directed edges on the path from the initial cognitive state node to the target cognitive state node to obtain the total path cost. For all cognitive state nodes in the reachable cognitive state node set, the path with the minimum total cost covering all cognitive state nodes in the reachable cognitive state node set is calculated as a candidate task execution sequence. Based on the mapping relationship between knowledge transfer source nodes and knowledge transfer target nodes recorded in the knowledge transfer interface, the cognitive switching cost between adjacent cognitive state nodes in the candidate task execution sequence is adjusted. The candidate task execution sequence with the minimum adjusted total cost is determined as the task execution sequence of the participant.

7. The method according to claim 1, characterized in that, Based on the participants' historical task execution efficiency characteristics, the task execution sequence is aligned with a time window to generate task allocation instructions, including: Extract the set of necessary knowledge nodes and task complexity features corresponding to each cognitive state node from the task execution sequence, and predict the expected execution time of the cognitive state node based on the set of necessary knowledge nodes, the task complexity features and the historical task execution efficiency features; The cognitive switching cost between adjacent cognitive state nodes in the task execution sequence is converted into a time buffer. The start time of the available time window is used as the time starting point, and execution time periods are allocated to the cognitive state nodes in the task execution sequence in sequence within the duration of the available time window. When the execution time of a cognitive state node in the task execution sequence exceeds the duration of the available time window, a subset of cognitive state nodes that can be delayed in execution is identified from the task execution sequence, and the subset of cognitive state nodes that can be delayed in execution is mapped to the next available time window of the participant. Based on the task execution sequence, the time period corresponding to each cognitive state node and the time window mapping relationship between the subset of cognitive state nodes that can be delayed, task allocation instructions are generated for the meeting participants.

8. A task adaptive allocation system for intelligent meeting collaboration, used to implement the method as described in any one of claims 1-7, characterized in that, include: The dependency resolution unit is used to construct a knowledge matching network and deduce the knowledge propagation path based on the knowledge coverage relationship between the task knowledge dependency features of the meeting and the knowledge graph features of the meeting participants. It identifies the necessary knowledge nodes and knowledge gap nodes for task execution, calculates the knowledge complementarity index among participants, and obtains the knowledge dependency resolution results. The granularity reconstruction unit is used to reconstruct the collaborative granularity of the task set based on the knowledge dependency parsing results and task collaboration granularity characteristics through a two-way adjustment mechanism of task decomposition and aggregation. When the number of knowledge gap nodes exceeds the knowledge coverage capacity of a single participant, the current task is decomposed into multiple sub-tasks and a knowledge transfer interface between sub-tasks is established according to the dependency order of the necessary knowledge nodes. The task aggregation unit is used to aggregate multiple tasks into a collaborative task group and establish a knowledge sharing mechanism when the overlap of the required knowledge nodes of multiple tasks exceeds a preset overlap threshold and the knowledge complementarity index of the corresponding participants exceeds a preset complementarity threshold, thereby obtaining the task structure after the collaborative granularity reconstruction. The task allocation unit is used to construct a cognitive state transition diagram based on the task structure reconstructed by the collaborative granularity and the cognitive switching cost characteristics of the participants, plan a task execution sequence with minimum cognitive switching cost for each participant in the cognitive state transition diagram based on the knowledge transfer interface, align the task execution sequence with a time window based on the historical task execution efficiency characteristics of the participants, and generate a task allocation instruction.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images