A composite knowledge crowdsourcing model and incentive model construction method

Abstract

The present invention discloses a composite knowledge crowdsourcing model and an incentive model construction method, comprising: firstly, dividing the crowdsourcing stages according to the project goals of the knowledge graph construction, and determining the task goals and capability requirements; then stratifying the crowdsourcing objects, clarifying the capability list and incentive preferences; then constructing a full-process three-layer mapping crowdsourcing model; finally, constructing an incentive model, realizing refined incentives through the three-layer mapping model, and improving task allocation and incentive efficiency; utilizing the dynamic resource allocation and reputation value update strategy of the incentive model to improve the adaptability and effectiveness of the incentive mechanism, while taking into account the heterogeneity of participants and the dynamic changes of tasks to ensure optimal output, thereby significantly improving the efficiency and quality of knowledge crowdsourcing task completion in the domain knowledge graph project.

Description

Technical Field

[0001] The present invention belongs to computer technology, specifically to the field of domain knowledge and model optimization technology, and in particular to a composite knowledge crowdsourcing model and an incentive model construction method. Background Art

[0002] With the rapid development of artificial intelligence (AI) technology, demand for multi-domain, long-term, and structured knowledge in specialized data is growing across various fields. High-quality multimodal knowledge graphs are the data foundation, supporting the expansion of the precision, depth, and breadth of modern knowledge services. Currently, the construction of knowledge graphs or knowledge bases in various specialized fields is a common task in important research projects at universities. Key stages such as knowledge system design, data collection and annotation, knowledge model design, and knowledge graph quality control require the in-depth involvement of knowledge crowdsourcing to improve the speed and scale of knowledge graph construction. Knowledge crowdsourcing for knowledge graphs primarily targets university groups, including expert faculty, graduate and doctoral students, and undergraduate students with a certain level of domain knowledge literacy. Compared to traditional crowdsourcing tasks, knowledge crowdsourcing is limited by project funding, completion timelines, and the high level of knowledge literacy and diverse incentive requirements of the crowdsourced audience. To more efficiently complete crowdsourcing tasks, it is necessary to design targeted crowdsourcing models and incentive structures that cater to the incentive needs of different groups.

[0003] At present, common knowledge crowdsourcing models mainly include open knowledge crowdsourcing model, internal knowledge crowdsourcing model, competitive knowledge crowdsourcing model, and user-segmented knowledge crowdsourcing model, as follows: ① Open knowledge crowdsourcing model: mobilize a wide range of professional groups through open platforms to complete tasks such as knowledge collection, organization, annotation and verification; participants come from a wide range of sources, rely on open collaboration and information sharing to improve the efficiency and quality of knowledge construction, and perform outstandingly in promoting interdisciplinary scientific research exchanges and knowledge sharing; ② Internal knowledge crowdsourcing model: takes enterprises or organizations as the main participants, while protecting internal information and maintaining competitive advantages, it gathers the wisdom of internal members to promote the enterprise Crowd intelligence innovation is suitable for tasks that require high confidentiality, taking into account both information security and efficiency improvement; ③ Competitive knowledge crowdsourcing model: through task decomposition and capability matching models, crowdsourcing tasks are assigned to different groups of participants, and participants are encouraged to provide high-quality solutions through a competition mechanism, emphasizing the matching of individual capabilities with task requirements, and is suitable for complex tasks that require innovative solutions; ④ User-segmented knowledge crowdsourcing model: for specific types of knowledge production tasks, users are segmented according to the quality and quantity of their contributions, and differentiated incentive strategies are implemented; with users as the core, it focuses on improving the quality and quantity of participants' output, and is suitable for knowledge production projects in specific fields.

[0004] At present, common crowdsourcing incentive models mainly include honor incentives that conduct reputation evaluation or ranking rewards for participants' contributions to enhance social recognition and sense of honor; emotional incentives that use social influence, interesting design or personalized measures to stimulate participants' emotional resonance and sense of belonging, reduce task costs and increase participation rates; interest incentives that increase the attractiveness of tasks through direct material rewards or combined with virtual rewards; and mechanism incentives that solve problems such as malicious bidding, free riding or unbalanced participation by optimizing the incentive distribution mechanism.

[0005] However, the above existing knowledge crowdsourcing models and incentive models have the following shortcomings in their application:

[0006] ① Lack of fine-grained analysis of the characteristics of heterogeneous groups: Existing studies mostly focus on a single type of crowdsourcing object, and do not fully analyze the differences in knowledge background, ability characteristics and task adaptation among different groups, and cannot meet the needs of complex group participation tasks; ② Insufficient targeting of incentive mechanism design: General incentive models fail to combine the specific crowdsourcing task context with the motivations and preferences of participants, lack fine-grained incentive design, and are difficult to effectively stimulate the enthusiasm of heterogeneous groups, resulting in significant differences in incentive effects among different groups; ③ Insufficient dynamic application and actual effect verification: Existing studies focus on the construction and verification of theoretical models, but lack in-depth research on participant behavioral feedback, dynamic changes in incentive effects, and long-term impacts in actual applications, affecting the reliability and feasibility of the model in practice. Summary of the Invention

[0007] In order to solve the above problems existing in the prior art, the present invention provides a composite knowledge crowdsourcing model and an incentive model construction method, comprising the following steps:

[0008] Step 1: Divide the knowledge graph construction project into crowdsourcing stages according to the overall goal and task logic, and determine the crowdsourcing stage task goals and crowdsourcing stage capability requirements according to the crowdsourcing stages;

[0009] Step 2: Stratify the knowledge crowdsourcing subjects according to their educational background to obtain multiple crowdsourcing participant groups, and determine the ability list of the crowdsourcing participant groups based on the crowdsourcing participant groups; conduct an incentive questionnaire survey on the crowdsourcing participant groups to determine their incentive preferences;

[0010] Step 3: Take the crowdsourcing stage task objectives as the first layer, each crowdsourcing participant group as the second layer, and the incentive method as the third layer; map the crowdsourcing stage task objectives to each crowdsourcing participant group based on the crowdsourcing participant group capability list, and map the crowdsourcing participant group to the corresponding incentive method based on the crowdsourcing participant group incentive preference to obtain a full-process three-layer mapping crowdsourcing model;

[0011] Step 4: Build an incentive model and apply it to the full-process three-layer mapping crowdsourcing model to crowdsource the knowledge graph construction project.

[0012] Furthermore, the crowdsourcing phase includes: ontology design, data source selection, design of available knowledge schemas in data sources, data cleaning and knowledge organization, and knowledge quality control.

[0013] Furthermore, the crowdsourcing participants include undergraduates, masters, doctors and expert teachers.

[0014] Furthermore, incentive methods include: material incentives, emotional incentives, honor incentives, opportunity incentives and ability development incentives.

[0015] Furthermore, the incentive model includes: output calculation of the crowdsourcing participant group, objective function, total resource constraints, upper and lower limit constraints on resource allocation, and reputation value update mechanism.

[0016] Furthermore, the output calculation formula of the crowdsourcing participant group is:

[0017]

[0018] in, represents the output of crowdsourcing participant group i, represents the proportion weight of crowdsourcing participant group i, represents the output coefficient of crowdsourcing participant group i, represents the weight preference of crowdsourcing participant group i for incentive method j, It represents the amount of resources allocated to crowdsourcing participant group i in terms of incentive method j, δ is a fluctuation term, 0.9<δ<1.1, represents the reputation value of crowdsourcing participant group i, represents the maximum value of the reputation value, λ is the long-term decay factor, t is the number of iterations, The burnout factor.

[0019] Furthermore, the objective function is:

[0020]

[0021] Among them, Z represents the total output of each iteration.

[0022] Furthermore, the total resource constraint is:

[0023]

[0024] in, X is the total amount of resources for each iteration, N The maximum value of the total amount of resources for each iteration.

[0025] Furthermore, the upper and lower bounds of resource allocation are:

[0026]

[0027]

[0028] in, is the remaining resources of crowdsourcing participant group i, is the remaining number of iterations, A The first constraint parameter for resource allocation upper and lower limits, B The second constraint parameter for resource allocation upper and lower limits, It represents the amount of resources allocated to crowdsourcing participant group i in terms of incentive method j.

[0029] Furthermore, the reputation value update mechanism is:

[0030]

[0031] in, is the reputation value of crowdsourcing participant group i, is the current output of crowdsourcing participant group i, is the output of crowdsourcing participant group i in the previous iteration, is the credit value increase adjustment coefficient, is the credit value decrease adjustment coefficient, is the rate of change of the output of the crowdsourcing participant group, It is the standard value of reputation value, min means the minimum value, and max means the maximum value.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] The present invention adopts a three-layer mapping model design to carry out refined incentives for different crowdsourcing participating groups, ensuring efficient task allocation and incentives; by constructing an incentive model to realize dynamic resource allocation and reputation value update strategy, the incentive resource allocation is adjusted in real time, the adaptability and effectiveness of the incentive mechanism are improved, and the enthusiasm of participants is avoided due to unreasonable incentive mechanism, thereby improving the efficiency and quality of task completion; taking into account the heterogeneity of participants and dynamic changes in the task process, the incentive strategy is adjusted in real time to ensure optimal output; it significantly improves the completion efficiency and quality of knowledge crowdsourcing tasks in domain knowledge graph construction projects, has strong adaptability, and can cope with complex and changeable practical application scenarios.

[0034] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1This is a flow chart of a composite knowledge crowdsourcing model and incentive model construction method provided by an embodiment of the present invention;

[0036] Figure 2 This is a three-layer crowdsourcing model diagram of the entire process of a multimodal agricultural knowledge graph construction project provided by an embodiment of the present invention;

[0037] Figure 3 This is a simulation curve diagram of the change in reputation value of a crowdsourcing participant group provided by an embodiment of the present invention;

[0038] Figure 4 This is a statistical comparison chart of the total output under the composite knowledge crowdsourcing model provided by an embodiment of the present invention and the total output of other models. DETAILED DESCRIPTION

[0039] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the scheme according to the present invention is described in detail below with reference to the accompanying drawings and specific implementation methods.

[0040] The aforementioned and other technical contents, features, and effects of the present invention are clearly presented in the following detailed description of the specific embodiments in conjunction with the accompanying drawings. Through the description of the specific embodiments, a deeper and more specific understanding of the technical means and effects adopted by the present invention to achieve the intended purpose can be obtained. However, the accompanying drawings are provided for reference and illustration purposes only and are not intended to limit the technical solutions of the present invention.

[0041] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations are intended to cover non-exclusive inclusion, such that an article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the article or device comprising the element.

[0042] like Figure 1 As shown, a composite knowledge crowdsourcing model and incentive model construction method provided by an embodiment of the present invention includes the following steps:

[0043] Step 1: Divide the knowledge graph construction project into crowdsourcing stages according to the overall goal and task logic, and determine the crowdsourcing stage task goals and crowdsourcing stage capability requirements according to the crowdsourcing stages.

[0044] The knowledge graph construction project is divided into five phases based on the overall goals and task logic: ontology design, data source selection, design of available knowledge models within the data source, data cleaning and knowledge organization, and knowledge quality control. By clarifying the core tasks and objectives of each phase, related tasks are grouped into the same phase, ensuring clear objectives and measurable outcomes for each phase while facilitating collaboration and division of labor.

[0045] Through a combination of interviews and actual needs analysis, including phased task objectives, key skill requirements, and commonly used tools, we clarified the crowdsourcing phase objectives and the actual crowdsourcing capability requirements required to achieve them. This phased division of the project and the rational definition of capability requirements provided important support for efficient project organization and collaboration. This not only clarified the task objectives but also ensured a good match between the task objectives and capability requirements, contributing to the smooth progress of the project and improved quality of the results.

[0046] As shown in Table 1, this embodiment provides a crowdsourcing stage, crowdsourcing stage task objectives, and crowdsourcing stage capability requirements for a multimodal agricultural knowledge graph construction project.

[0047] Table 1 Task objectives and capability requirements for the knowledge crowdsourcing phase of the multimodal agricultural knowledge graph construction project

[0048]

[0049] Step 2: Stratify the knowledge crowdsourcing objects according to their educational background to obtain multiple crowdsourcing participant groups, and determine the ability list of the crowdsourcing participant groups based on the crowdsourcing participant groups.

[0050] The knowledge crowdsourcing objects constructed in the knowledge graph are stratified according to their common educational background, and the knowledge crowdsourcing participants are divided into expert teachers, doctoral students, master's students and undergraduates.

[0051] Through interviews, we collected the ability performance of five types of crowdsourcing participants, discussed and sorted out the specific abilities of the participants, and obtained a list of the abilities of the crowdsourcing participants. That is, the abilities of the crowdsourcing participants were summarized as follows: undergraduates were responsible for basic collection and preprocessing, master's students participated in data analysis and architecture design, doctoral students conducted standardized design and knowledge reasoning, and expert teachers provided field guidance and design support.

[0052] As shown in Table 2, this embodiment provides a list of crowdsourcing participant groups and capabilities of crowdsourcing participant groups for a multimodal agricultural knowledge graph construction project. The knowledge crowdsourcing participant groups of this project are divided into five categories: undergraduates, master's students, doctoral students, and expert teachers (including agricultural experts and information science experts).

[0053] Table 2 List of crowdsourcing participants and their capabilities in the multimodal agricultural knowledge graph construction project

[0054]

[0055] Conduct a questionnaire survey on incentive methods for crowdsourcing participants to determine their incentive preferences:

[0056] A 15-question questionnaire was designed for the crowdsourcing participants of the multimodal agricultural knowledge graph construction project. The questionnaire used a combination of multiple-choice questions, ranking questions, scoring questions (using a five-level scale) and open-ended questions. The multiple-choice questions, ranking questions, scoring questions and open-ended questions in the questionnaire were processed using a proportional weighting method to determine the incentive preferences of the crowdsourcing participants. The incentive methods included: material incentives, emotional incentives, honor incentives, opportunity incentives and ability development incentives.

[0057] Step 3: Take the task objectives of the crowdsourcing stage as the first layer, each crowdsourcing participant group as the second layer, and the incentive method as the third layer; map the task objectives of the crowdsourcing stage to each crowdsourcing participant group according to the capability list of the crowdsourcing participant group, and map the crowdsourcing participant group to the incentive method according to the incentive preference of the crowdsourcing participant group to obtain a full-process three-layer mapping crowdsourcing model.

[0058] like Figure 2 As shown, a full-process three-layer mapping crowdsourcing model for a multimodal agricultural knowledge graph construction project is provided by an embodiment of the present invention. The five crowdsourcing stages (S1 to S5) and their subtasks in the multimodal agricultural knowledge graph construction project are gradually connected with different types of crowdsourcing participant groups and incentive methods through a mapping method. The crowdsourcing stage task goals are taken as the first layer, each crowdsourcing participant group is taken as the second layer, and the incentive method is taken as the third layer. That is, the crowdsourcing stage task goals are mapped to each crowdsourcing participant group according to the crowdsourcing participant group capability list, and the crowdsourcing participant group is mapped to the incentive method according to the crowdsourcing participant group incentive preference to obtain a full-process three-layer mapping crowdsourcing model. The heterogeneity of the crowdsourcing participant group is introduced through the full-process three-layer mapping crowdsourcing model, so that the composite crowdsourcing model has both adaptability and incentive.

[0059] The S1 ontology design stage includes S1.1 designing the top-level ontology and domain ontology in the agricultural field and S1.2 designing the application ontology in the agricultural field. This stage requires the crowdsourcing objects to have professional agricultural knowledge and ontology design experience to ensure the coverage, accuracy and practicality of the ontology.

[0060] The S2 data source selection stage includes S2.1 selection of open source authoritative agricultural websites, S2.2 selection of closed source professional data sources, and S2.3 literature selection. This stage requires the crowdsourcing objects to have the ability to identify agricultural data and screen data sources to ensure the reliability, coverage, and timeliness of the data sources.

[0061] The stage of designing available knowledge schemas in the S3 data source includes S3.1 data source availability assessment and S3.2 agricultural knowledge graph coarse-grained triple schema design. This stage requires crowdsourcing objects to have data source analysis and knowledge organization capabilities to ensure that the knowledge organization of crowdsourcing data is reasonable.

[0062] The S4 data cleaning and knowledge organization stage includes S4.1 data cleaning of crawled data, S4.2 data review, S4.3 data organization and fine-grained triple schema design, and S4.2 data storage. This stage requires the crowdsourcing objects to have data processing and review capabilities, and improve the overall quality and availability of data through standardized operations.

[0063] The S5 knowledge quality control stage includes S5.1 discovery and correction of erroneous knowledge, S5.2 discovery and completion of missing knowledge, and S5.3 updating of expired knowledge. This stage requires the crowdsourcing objects to have strong knowledge evaluation and management capabilities, and ensure the accuracy and completeness of the knowledge graph through strict quality control and knowledge reasoning.

[0064] The task objectives of the crowdsourcing stage are mapped to the crowdsourcing participant groups (O1 to O5) through connecting lines: O1 undergraduates, O2 master's students, O3 doctoral students, O4 agricultural experts, and O5 information science experts; the incentive preferences and intensities of the crowdsourcing participant groups are mapped to different incentive methods (I1 to I5) using connecting lines of different thicknesses, among which the intensity of incentive preferences with thick lines is high and the intensity of incentive preferences with thin lines is low; the incentive methods include: I1 material incentives, I2 emotional incentives, I3 honor incentives, I4 opportunity incentives, and I5 ability development incentives.

[0065] The three-layer mapping relationship clearly demonstrates the roles and responsibilities of crowdsourcing participants with different professional backgrounds and knowledge levels at each stage, and provides a basis for selecting appropriate incentive methods, thereby supporting the efficient implementation of the multimodal agricultural knowledge graph construction project.

[0066] Step 4: Build an incentive model and apply it to the full-process three-layer mapping crowdsourcing model to crowdsource the knowledge graph construction project.

[0067] The incentive model includes: output calculation of the crowdsourcing participant group, objective function, total resource constraints, upper and lower limit constraints on resource allocation, and reputation value update mechanism.

[0068] The output calculation formula of the crowdsourcing participant group is:

[0069]

[0070] in, represents the output of crowdsourcing participant group i, represents the proportion weight of crowdsourcing participant group i, represents the output coefficient of crowdsourcing participant group i, represents the weight preference of crowdsourcing participant group i for incentive method j, It represents the amount of resources allocated to crowdsourcing participant group i in terms of incentive method j, δ is a fluctuation term, 0.9<δ<1.1, represents the reputation value of crowdsourcing participant group i, represents the maximum value of the reputation value, λ is the long-term decay factor, t is the number of iterations, The burnout factor.

[0071] in, The proportional weighting method is used to determine the proportional weight according to the proportion of personnel in the multimodal agricultural knowledge graph construction project in this embodiment. The sum of the weights is 1, and r1 to r5 correspond to the respective proportional weights of undergraduates, masters, doctors, agricultural experts, and information science experts.

[0072] The specific values ​​are determined based on the participation allowances (or wages) and time invested by the crowdsourcing participants in the multimodal agricultural knowledge graph construction project in this embodiment. The allowances (or wages) for undergraduates, master's students, doctoral students, agricultural experts, and information science experts are 500 yuan / month, 1200 yuan / month, 3200 yuan / month, 15,000 yuan / month, and 15,000 yuan / month, respectively. Since agricultural experts spend less time participating, their allowances (or wages) are reduced to one-third when calculating. , c1 to c5 correspond to the output coefficients of undergraduates, masters, doctors, agricultural experts and information science experts respectively.

[0073] The weight of each incentive method in the ranking question is obtained by the crowdsourcing participant group in the multimodal agricultural knowledge graph construction project in this embodiment. The weight is multiplied by the average score of the corresponding scoring question. Finally, the proportional weighting method is used to obtain the incentive preference of each crowdsourcing participant group, that is, the preference weight matrix of each crowdsourcing participant group i for different incentive methods j .

[0074] and It reflects the basic performance capabilities of the crowdsourcing participant groups, resulting in differentiated output capabilities. By combining these coefficients with resources and reputation values, the incentive model can dynamically adapt to different group characteristics, ensuring that output not only depends on resource input, but also takes into account the influence of the inherent characteristics of each group, ensuring that the incentive mechanism adapts to the diverse needs of crowdsourcing participants; under the condition of limited resources, by combining resources, reputation values ​​and group characteristics, the optimal allocation of resources is achieved to maximize overall output while ensuring fairness, providing a scientific output evaluation basis for the incentive model.

[0075] Through according to Exponentiation simulates the phenomenon of diminishing marginal effects of resources, that is, as resource allocation increases, the output growth it brings gradually decreases, reflecting the nonlinear contribution characteristics of resources, that is, simply increasing resources may not necessarily bring about proportional output growth, thereby making the incentive model more reasonable.

[0076] The reputation value of crowdsourcing participant group i As a variable, the incentive model is introduced to update the reputation value in each iteration so that the group with higher reputation value can produce higher output under the same resource conditions. It ensures the regulatory effect of reputation value on output efficiency; by assuming that groups with high reputation value have stronger resource utilization capabilities, it further motivates participants to maintain a high reputation level, thereby promoting long-term stability and high-quality output.

[0077] The long-term attenuation factor λ decreases over time, reflecting the natural adaptability of the crowdsourcing participant group to the incentive method in long-term tasks. t is the number of iterations, α is the attenuation constant that controls the attenuation rate, and is calculated by fitting the attenuation ratio. e is the base of the exponential function and is set to 2.71828. The calculation formula for the long-term attenuation factor λ is as follows:

[0078]

[0079] Burnout Factor It is used to measure the output fatigue level of crowdsourcing participant group i in the recent time window, identify signs of fatigue in high-frequency or high-load tasks, and dynamically reduce their output influence, thereby effectively protecting the group's participation enthusiasm and alleviating the side effects of over-incentives; represents the number of times the crowdsourcing object participates in group i to produce fatigue results, t is the number of iterations, and k represents iterations, max represents the maximum number of times the crowdsourcing object participates in group i to produce fatigue results, represents the output of crowdsourcing participant group i; in this embodiment, the most recent time window is 15 days, and the output is calculated twice a week. Therefore, the number of fatigue results produced by crowdsourcing participant group i is taken as 5, burnout factor The calculation formula is as follows:

[0080]

[0081] The objective function is:

[0082]

[0083] Among them, Z represents the total output of each iteration, Represents the output of crowdsourcing participant group i.

[0084] The total resource constraint is:

[0085]

[0086] in, X is the total amount of resources for each iteration, represents the amount of resources allocated to crowdsourcing participant group i in terms of incentive method j, N The maximum value of the total amount of resources for each iteration. In this embodiment, N is set to 1000. It can be seen that the total amount of resources for the project is the total amount of resources for each iteration multiplied by the total number of iterations. The iteration stopping condition is that the total number of iterations is reached.

[0087] The upper and lower bounds of resource allocation are:

[0088]

[0089]

[0090] in, is the remaining resources of crowdsourcing participant group i, is the remaining number of iterations, represents the amount of resources allocated to crowdsourcing participant group i in terms of incentive method j, A The first constraint parameter for resource allocation upper and lower limits, B The second constraint parameter for resource allocation upper and lower limits, represents the amount of resources allocated to crowdsourcing participant group i in terms of incentive method j; in this embodiment, A is 0.7, B is 1.3; the amount of resources allocated by any incentive method to any crowdsourcing participant group is greater than or equal to 0, that is, .

[0091] The reputation value update mechanism is:

[0092]

[0093] in, is the reputation value of crowdsourcing participant group i, is the current output of crowdsourcing participant group i, is the output of crowdsourcing participant group i in the previous iteration, is the credit value increase adjustment coefficient, is the credit value decrease adjustment coefficient, is the rate of change of the output of the crowdsourcing participant group, It is the standard value of reputation value, min means the minimum value, and max means the maximum value.

[0094] The reputation value reflects the performance of the crowdsourcing participant group and affects its output level. A high reputation value represents high reliability. Through this reputation value update mechanism, the model further encourages crowdsourcing participants to continuously improve their work performance to ensure the long-term stability and efficient implementation of the crowdsourcing project.

[0095] The present invention designs a Python program to implement the incentive model and conducts simulation experiments on the PyCharm platform. The multimodal agricultural knowledge graph is used to build a simulation setting. The duration of the knowledge crowdsourcing project is set to three years, i.e. 152 weeks. Reports are set twice a week, and a total of 312 iterations are performed. The dynamic changes in resource allocation, output, and reputation value in the project are analyzed. Figure 3 , which is a simulation curve diagram of the change in reputation value of a crowdsourcing participant group provided by an embodiment of the present invention.

[0096] The simulation results show that although there are certain fluctuations in resource allocation in each iteration, the resource allocation within each group is always consistent with the weight of the group in the corresponding incentive method, indicating that there is a stable correspondence between resource allocation and incentive preferences in the process of maximizing group output. The incentive model of the present invention can optimize resource allocation according to group incentive preferences and thus enhance the enthusiasm of participants.

[0097] During the simulation, although the reputation value fluctuated at certain moments, the overall trend showed a clear upward trend, indicating that despite the personnel flow in the knowledge crowdsourcing project, the overall reputation value continued to rise, which is consistent with the actual situation of the agricultural knowledge crowdsourcing project. It also reflects that in the long-term task cooperation, the trust relationship between the contractor and the contractor has gradually strengthened. The increase in reputation value not only reduces the transaction costs within the organization, but also creates a harmonious collaborative environment and improves the overall efficiency of the project.

[0098] In addition, to verify the superiority of the incentive model, the total output of the incentive model in 312 iterations was compared with the total output of five static resource allocation schemes; the five static resource allocation schemes use the same output calculation formula but do not perform iterative adjustments. The specific allocation methods of the five static resource allocation schemes are as follows:

[0099] Distribute the total resources of each round evenly among all groups and incentive methods; distribute the total resources of each round according to the proportion of group population, and then distribute them evenly within the group; distribute the total resources of each round according to the proportion of group population, and then distribute them within the group according to the incentive preference weight; distribute the total resources of each round according to the average contribution of the project, and then distribute them evenly within the group; distribute the total resources of each round according to the average contribution of the project, and then distribute them within the group according to the incentive preference weight.

[0100] like Figure 4 As shown in the figure, it is a statistical comparison chart of the total output under the composite knowledge crowdsourcing model provided by the embodiment of the present invention and the total output of other models. It can be seen that the iterative resource allocation corresponding to the method of the present invention significantly optimizes the total output while ensuring that the minimum output target of each group is achieved, and its performance is better than the other five static resource allocation schemes.

[0101] The present invention adopts a three-layer mapping model design to carry out refined incentives for different crowdsourcing participating groups, ensuring efficient task allocation and incentives; by constructing an incentive model to realize dynamic resource allocation and reputation value update strategy, the incentive resource allocation is adjusted in real time, and the adaptability and effectiveness of the incentive mechanism are improved, avoiding the decline in participants' enthusiasm due to unreasonable incentive mechanism, thereby improving the efficiency and quality of task completion; it also takes into account the heterogeneity of participants and dynamic changes in the task process, and adjusts the incentive strategy in real time to ensure optimal output; it significantly improves the completion efficiency and quality of knowledge crowdsourcing tasks in domain knowledge graph projects, has strong adaptability, and can cope with complex and changeable practical application scenarios.

[0102] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A composite knowledge crowdsourcing model and incentive model construction method, characterized by: The following steps are involved: Step 1: Divide the knowledge graph construction project into crowdsourcing stages according to the overall goal and task logic, and determine the crowdsourcing stage task goals and crowdsourcing stage capability requirements according to the crowdsourcing stages; Step 2: Stratify the knowledge crowdsourcing subjects according to their educational background to obtain multiple crowdsourcing participant groups, and determine the ability list of the crowdsourcing participant groups based on the crowdsourcing participant groups; conduct an incentive questionnaire survey on the crowdsourcing participant groups to determine their incentive preferences; Step 3: Take the crowdsourcing stage task objectives as the first layer, each crowdsourcing participant group as the second layer, and the incentive method as the third layer; map the crowdsourcing stage task objectives to each crowdsourcing participant group based on the crowdsourcing participant group capability list, and map the crowdsourcing participant group to the corresponding incentive method based on the crowdsourcing participant group incentive preference to obtain a full-process three-layer mapping crowdsourcing model; Step 4: Construct an incentive model and apply it to the crowdsourcing of knowledge graph construction projects in the full-process three-layer mapping crowdsourcing model; the incentive model includes: output calculation of the crowdsourcing participant group, objective function, total resource constraints, upper and lower limit constraints on resource allocation, and reputation value update mechanism; The output calculation formula of the crowdsourcing participant group is: in, represents the output of crowdsourcing participant group i, represents the proportion weight of crowdsourcing participant group i, represents the output coefficient of crowdsourcing participant group i, represents the weight preference of crowdsourcing participant group i for incentive method j, It represents the amount of resources allocated to crowdsourcing participant group i in terms of incentive method j, δ is a fluctuation term, 0.9<δ<1.1, represents the reputation value of crowdsourcing participant group i, represents the maximum value of the reputation value, λ is the long-term decay factor, t is the number of iterations, The burnout factor is used to measure the output fatigue level of crowdsourcing participant group i in the recent time window and identify signs of fatigue under high-frequency or high-load tasks.

2. A composite knowledge crowdsourcing model and incentive model construction method according to claim 1, characterized in that: The crowdsourcing stage includes: ontology design, data source selection, design of available knowledge models in the data source, data cleaning and knowledge organization, and knowledge quality control.

3. A composite knowledge crowdsourcing model and incentive model construction method according to claim 1, characterized in that: The crowdsourcing participants include undergraduates, masters, doctors and expert teachers.

4. A composite knowledge crowdsourcing model and incentive model construction method according to claim 1, characterized in that: The incentive methods include: material incentives, emotional incentives, honor incentives, opportunity incentives and ability development incentives.

5. A composite knowledge crowdsourcing model and incentive model construction method according to claim 1, characterized in that: The objective function is: Among them, Z represents the total output of each iteration.

6. A composite knowledge crowdsourcing model and incentive model construction method according to claim 1, characterized in that: The total resource constraint is: in, X is the total amount of resources for each iteration, N The maximum value of the total amount of resources for each iteration.

7. A composite knowledge crowdsourcing model and incentive model construction method according to claim 1, characterized in that: The upper and lower bounds of resource allocation are: in, is the remaining resources of crowdsourcing participant group i, is the remaining number of iterations, A The first constraint parameter for resource allocation upper and lower limits, B The second constraint parameter is the upper and lower bounds for resource allocation.

8. A composite knowledge crowdsourcing model and incentive model construction method according to claim 1, characterized in that: The reputation value update mechanism is: in, is the current output of crowdsourcing participant group i, is the output of crowdsourcing participant group i in the previous iteration, is the credit value increase adjustment coefficient, is the credit value decrease adjustment coefficient, is the rate of change of the output of the crowdsourcing participant group, It is the standard value of reputation value, min means the minimum value, and max means the maximum value.

Images