Oral talent training method and device based on incentive mechanism and medium

Abstract

The invention discloses a talent training method and device based on an incentive mechanism and a medium. The method comprises the following steps: calculating according to learning data to obtain initial comprehensive incentive; obtaining feedback data of the user on the initial comprehensive excitation, and adjusting the excitation intensity of the initial comprehensive excitation according to the feedback data to obtain comprehensive excitation; generating a first improvement suggestion according to the learning data and the comprehensive excitation; establishing an improved suggestion set according to the behavior pattern and the change trend of the user in the learning data; a spoken training scheme is generated from the first improvement suggestion and the set of improvement suggestions. The invention provides the oral talent training method and device based on the incentive mechanism and the medium, the incentive is matched with the expectation and demand of the user by adjusting the incentive intensity, the sustainable implementation of the training scheme is ensured, suggestions in multiple aspects are generated according to the behavior and change of the user, and the training efficiency is improved. The problem that it is difficult to carry out personalized training on a user by dynamically adjusting a talent incentive mechanism based on the performance and progress of the user can be solved.

Description

Technical Field

[0001] The invention relates to the field of speech training, and in particular to an eloquence training method, device and medium based on an incentive mechanism. Background Art

[0002] In the field of eloquence training, incentive mechanisms can stimulate the trainees' internal motivation and self-confidence. Through incentive mechanisms, such as affirmation, encouragement and rewards, trainees can gradually overcome self-doubt and nervousness, dare to express their opinions in public, thereby enhancing self-confidence and improving eloquence performance; and eloquence training is a long-term process that requires trainees to constantly summarize and adjust. Incentive mechanisms can encourage trainees to persevere when encountering difficulties and have the courage to try new methods and strategies, thereby continuously improving their eloquence ability. The existing eloquence incentive mechanism is mainly to establish an eloquence training reward system during the training process, reward trainees with excellent performance and obvious progress, or publicly praise and recognize the excellent performance of trainees in the training team or on social platforms, which can stimulate the trainees' sense of pride and honor and further promote their enthusiasm.

[0003] However, existing eloquence training programs often adopt a one-size-fits-all approach in terms of incentive mechanisms, that is, the same reward standards and strategies are applied to all trainees. However, since each trainee has different eloquence foundation, learning speed, interest preferences and goals, this one-size-fits-all reward method is difficult to meet the needs of all trainees, thus reducing the incentive effect; and, since existing incentive mechanisms often lack feedback functions, it is difficult to adjust the reward strategy in a timely manner according to the trainees' actual performance and progress, which will lead to delayed and inaccurate rewards, making it difficult for trainees to feel their own progress and achievements, thereby reducing their motivation to learn. Summary of the invention

[0004] The present invention provides an eloquence training method, device and medium based on an incentive mechanism, so as to solve the problem that it is difficult to dynamically adjust the eloquence incentive mechanism based on the performance and progress of the user and to provide personalized training for the user.

[0005] Acquire the user's learning data of a preset course; wherein the preset course is established based on the user's learning goal and according to the terminal cost function;

[0006] Calculate an initial comprehensive incentive based on the learning progress of the learning data;

[0007] Acquiring feedback data of the user on the initial comprehensive incentive, and adjusting the incentive intensity of the initial comprehensive incentive according to the feedback data to obtain a comprehensive incentive;

[0008] generating a first improvement suggestion according to the learning data and the comprehensive incentive;

[0009] Establishing a set of improvement suggestions based on the behavior patterns and change trends of users in the learning data;

[0010] An eloquence training program is generated based on the first improvement suggestion and the improvement suggestion set.

[0011] The present invention establishes preset courses based on the terminal cost function, which can ensure that the course design meets the preset quality standards and realize the optimization of the course design. The feedback data provides the system with a basis for improving the incentive strategy. By analyzing the feedback of the user, it can be identified which incentive measures are effective. Therefore, by adjusting the incentive intensity of the initial comprehensive incentive, it can be ensured that the incentive provided matches the user's expectations and needs. Since the learning data reflects the strengths, weaknesses and existing problems of the user in eloquence learning, the comprehensive incentive after adjusting the intensity not only considers the actual performance of the user, but also considers the user's psychological state and learning motivation. Therefore, the first eloquence improvement suggestion generated on this basis can stimulate the user's learning interest and motivation, so that the user can participate in eloquence learning more actively. By analyzing the user's behavior pattern and change trend in the learning data, the user's learning habits, points of interest and possible difficulties can be deeply understood. Based on these generated improvement suggestion sets, the user can be helped to locate the problem more accurately and provide effective solutions. Therefore, the eloquence training program generated by combining the first improvement suggestion and the improvement suggestion set not only makes a dynamic adjustment to the eloquence incentive machine, but also fully considers the user's eloquence advantages and disadvantages, thereby ensuring the effectiveness of the training effect.

[0012] Compared with the prior art, the present invention achieves matching of incentives with users' expectations and needs by adjusting the incentive intensity of the incentive mechanism, thereby ensuring the sustainability of the training program. It also generates various suggestions based on users' behavior patterns and changing trends to further match users' personal eloquence. Therefore, it can solve the problem of difficulty in dynamically adjusting the eloquence incentive mechanism based on users' performance and progress and providing personalized training for users.

[0013] As a preferred solution, the preset course is established based on the user's learning goal and according to the terminal cost function, specifically:

[0014] Establishing a loss function according to the user's learning goal and actual eloquence performance;

[0015] Establishing a basic cost function according to the gap between the target score and the actual evaluation score of the user, and establishing the terminal cost function by introducing a linear penalty term into the basic cost function;

[0016] The preset course is established according to the loss function and the terminal cost function.

[0017] This preferred solution quantifies the gap between the user's expected target score and the actual evaluation score achieved during the training process through the terminal cost function, so that the user can clearly understand what price will be faced if the preset goal is not achieved at the end of the training. This helps the user to always have a clear understanding of the goal during the training process; if the value of the cost function is high, it means that the gap between the training result and the goal is large, and the training method or strategy needs to be further adjusted. Combining the loss function and the terminal cost function to establish a preset course can provide clear guidance for the course design of eloquence training, determine which course elements or teaching methods are most effective in improving the user's eloquence skills, and thus design training courses in a targeted manner.

[0018] As a preferred solution, the initial comprehensive incentive is calculated according to the learning progress of the learning data, specifically:

[0019] According to the behavior pattern and its corresponding time node, identifying the user's behavior type from the learning data;

[0020] Based on the behavior type, a cumulative reward amount is calculated according to the additional reward obtained by the user for completing the preset learning task;

[0021] The initial comprehensive incentive is obtained by merging and calculating the accumulated reward amount and the behavior type.

[0022] This preferred solution helps to understand the user's learning habits, preferences and progress by identifying the user's learning behavior type. The initial comprehensive incentive is generated by integrating the cumulative reward amount and the behavior type, which not only takes into account the user's effort, but also takes into account the user's learning behavior type and learning style, so the initial comprehensive incentive can well meet the user's needs and expectations.

[0023] As a preferred solution, the initial comprehensive incentive is specifically:

[0024]

[0025] Among them, R(t) is the cumulative reward amount, P1(x,t) is the recognition probability of the behavior type, α and β are the coefficients for adjusting the static reward and the coefficients for adjusting the impact of dynamic behavior changes, t is time, and x is a single behavior type.

[0026] As a preferred solution, the excitation intensity of the initial comprehensive excitation is adjusted according to the feedback data to obtain a comprehensive excitation, specifically:

[0027] According to the change amount of user behavior in the feedback data and the average feedback score, adjusting the incentive intensity of the initial comprehensive incentive at a preset time to obtain a single-time incentive;

[0028] Adjusting the sensitivity of the initial comprehensive stimulus from the initial moment to the preset moment to obtain a comprehensive moment stimulus;

[0029] The comprehensive incentive is composed of the single moment incentive and the comprehensive moment incentive.

[0030] This preferred solution dynamically adjusts the intensity of the initial comprehensive incentive at a specific preset moment according to the change in user behavior and the average feedback score, which can make the incentive more in line with the user's current needs and status. In addition, the single-moment incentive, as a change in the intensity of the incentive at a specific moment, can achieve real-time adjustment of the incentive strategy. Adjusting the sensitivity of the initial comprehensive incentive from the initial moment to the preset moment means that it is possible to effectively pay attention to the trend and speed of the incentive change over time, ensure the continuity and consistency of the incentive plan in time, and avoid sudden changes in incentives from causing adverse effects on users; and the comprehensive moment incentive, as a way to adjust the overall incentive effect, can ensure the adaptability and sustainability of the incentive mechanism.

[0031] As a preferred solution, a set of improvement suggestions is established based on the behavior patterns and change trends of users in the learning data, specifically:

[0032] Capturing the skills of the learning data through a preset eloquence performance analysis model to obtain a comprehensive eloquence performance score;

[0033] Calculate the interaction index according to the interaction mode and interaction times of the users in the learning data;

[0034] By integrating the comprehensive score of eloquence performance and the interaction index, a comprehensive data analysis score is obtained;

[0035] Identify behavioral patterns and change trends in the learning data to obtain comprehensive analysis results;

[0036] The improvement suggestion set is established according to the comprehensive data analysis score and the comprehensive analysis result.

[0037] This preferred solution can provide a quantitative indicator for the user's eloquence performance by capturing the skills in the learning data to obtain a comprehensive score of eloquence performance, thereby clearly understanding the user's eloquence skill level; the interaction index is calculated based on the user's interaction method and interaction frequency in the learning data, reflecting the user's participation and enthusiasm in the learning process. Therefore, the comprehensive data analysis score obtained by fusion on this basis can provide a more comprehensive evaluation of the user's learning effect. According to the comprehensive score of eloquence performance and the comprehensive analysis results, a set of improvement suggestions is established. These suggestions can put forward specific improvement suggestions for the user's eloquence skills and interaction methods, helping users to improve their eloquence ability more effectively.

[0038] As a preferred solution, the improvement suggestion set is established according to the comprehensive data analysis score and the comprehensive analysis result, specifically:

[0039] Extracting information from the comprehensive data analysis score and the comprehensive analysis result, and obtaining an insight score by adjusting an offset of the output information;

[0040] Based on the insight score, an initial set of improvement suggestions is obtained according to probability calculation of given different suggestions;

[0041] Calculating an overall feedback score according to the insight score and the weight of each suggestion in the initial set of improvement suggestions;

[0042] According to the overall feedback score, several suggestions that meet a preset condition are selected from the initial set of improvement suggestions to obtain the set of improvement suggestions.

[0043] This preferred solution extracts information from the comprehensive data analysis score and comprehensive analysis results, which can capture key learning performance data, help to have a deeper understanding of the user's learning status, and discover potential improvement points, and then customize a highly targeted and practical initial improvement suggestion set based on the user's specific learning situation and potential problems. The concept of overall feedback score is introduced, which combines the insight score and the weight of each suggestion in the initial improvement suggestion set, which helps to more comprehensively understand the effect and feasibility of the initial suggestion set, and thus screen the initial improvement suggestion set, which means that suggestions that better meet the actual needs of users can be dynamically selected.

[0044] As a preferred solution, the eloquence performance analysis model is specifically:

[0045]

[0046] Among them, x k (s) is the real-time measurement value of the kth eloquence feature at time s, α k and β k are the attenuation factor and weight respectively, e is a natural constant, t is the total time, and t0 is the real time.

[0047] The present invention also provides an eloquence training device based on an incentive mechanism, comprising a data module, an incentive module, a feedback module, a suggestion module, a collection module and a solution module;

[0048] The data module is used to obtain the user's learning data on the preset course; wherein the preset course is established based on the user's learning goal and according to the terminal cost function;

[0049] The incentive module is used to calculate an initial comprehensive incentive according to the learning progress of the learning data;

[0050] The feedback module is used to obtain the user's feedback data on the initial comprehensive incentive, and adjust the incentive intensity of the initial comprehensive incentive according to the feedback data to obtain a comprehensive incentive;

[0051] The suggestion module is used to generate a first improvement suggestion based on the learning data and the comprehensive incentive;

[0052] The collection module is used to establish an improvement suggestion set according to the user's behavior pattern and change trend in the learning data;

[0053] The solution module is used to generate an eloquence training solution based on the first improvement suggestion and the improvement suggestion set.

[0054] As a preferred solution, the data module includes a loss unit, a gap unit and a course unit;

[0055] The loss unit is used to establish a loss function according to the user's learning goal and actual eloquence performance;

[0056] The gap unit is used to establish a basic cost function according to the gap between the target score and the actual evaluation score of the user, and to establish the terminal cost function by introducing a linear penalty term into the basic cost function;

[0057] The course unit is used to establish the preset course according to the loss function and the terminal cost function.

[0058] As a preferred solution, the incentive module includes a behavior unit, a reward unit and a fusion unit;

[0059] The behavior unit is used to identify the user's behavior type from the learning data according to the behavior pattern and its corresponding time node;

[0060] The reward unit is used to calculate the cumulative reward amount based on the behavior type and the additional reward obtained by the user for completing the preset learning task;

[0061] The fusion unit is used to obtain the initial comprehensive incentive by performing a fusion calculation on the accumulated reward amount and the behavior type.

[0062] As a preferred solution, the initial comprehensive incentive is specifically:

[0063]

[0064] Among them, R(t) is the cumulative reward amount, P1(x,t) is the recognition probability of the behavior type, α and β are the coefficients for adjusting the static reward and the coefficients for adjusting the impact of dynamic behavior changes, t is time, and x is a single behavior type.

[0065] As a preferred solution, the feedback module includes a single unit, a comprehensive unit and a constituent unit;

[0066] The single unit is used to adjust the incentive intensity of the initial comprehensive incentive at a preset time according to the change amount of user behavior in the feedback data and the average feedback score to obtain a single-time incentive;

[0067] The synthesis unit is used to adjust the sensitivity of the initial comprehensive excitation from the initial moment to the preset moment to obtain a comprehensive moment excitation;

[0068] The constituent unit is used to constitute the comprehensive excitation from the single moment excitation and the comprehensive moment excitation.

[0069] As a preferred solution, the collection module includes a presentation unit, an interaction unit, an analysis unit, a change unit and a collection unit;

[0070] The performance unit is used to perform skill capture on the learning data through a preset eloquence performance analysis model to obtain a comprehensive score of eloquence performance;

[0071] The interaction unit is used to calculate the interaction index according to the interaction mode and interaction times of the users in the learning data;

[0072] The analysis unit is used to obtain a comprehensive data analysis score by integrating the comprehensive score of eloquence performance and the interaction index;

[0073] The change unit is used to identify the behavior pattern and change trend in the learning data to obtain a comprehensive analysis result;

[0074] The collection unit is used to establish the improvement suggestion set according to the comprehensive data analysis score and the comprehensive analysis result.

[0075] As a preferred solution, the set unit includes the insight subunit, the probability subunit, the weight subunit and the improvement subunit;

[0076] The insight subunit is used to extract information from the comprehensive data analysis score and the comprehensive analysis result, and obtain an insight score by adjusting the offset of the output information;

[0077] The probability subunit is used to obtain an initial set of improvement suggestions based on the insight score and according to the probability calculation of different given suggestions;

[0078] The weight subunit is used to calculate the overall feedback score according to the insight score and the weight of each suggestion in the initial improvement suggestion set;

[0079] The improvement subunit is used to select several suggestions that meet preset conditions in the initial improvement suggestion set according to the overall feedback score to obtain the improvement suggestion set.

[0080] As a preferred solution, the eloquence performance analysis model is specifically:

[0081]

[0082] Among them, x k (s) is the real-time measurement value of the kth eloquence feature at time s, α k and β k are the attenuation factor and weight respectively, e is a natural constant, t is the total time, and t0 is the real time.

[0083] The present application also provides a storage medium, on which a computer program is stored. The computer program is called and executed by a computer to implement the above-mentioned eloquence training method based on the incentive mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Figure 1 It is a flowchart of an eloquence training method based on an incentive mechanism provided in an embodiment of the present application;

[0085] Figure 2 It is a structural schematic diagram of an eloquence training device based on an incentive mechanism provided in an embodiment of the present application. DETAILED DESCRIPTION

[0086] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0087] In the description of this application, it should be understood that the term "first" is used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "several" means two or more.

[0088] The embodiment of the present application provides an eloquence training method based on an incentive mechanism, which is mainly used in situations where it is necessary to dynamically adjust the intensity of the eloquence incentive mechanism according to the user's performance and progress, and generate a personalized eloquence training plan based on the user's learning data to train the user's eloquence.

[0089] Embodiment 1:

[0090] See also Figure 1 The embodiment of the present application provides an eloquence training method based on an incentive mechanism, including S1 to S6, and the specific implementation steps are as follows:

[0091] S1. Obtain the user's learning data on preset courses; wherein the preset courses are based on the user's learning goals and are established according to the terminal cost function.

[0092] Step S1 of the embodiment of the present application includes S1.1 to S1.2; wherein S1.1 is the process of establishing a preset course, and S1.2 is the process of obtaining learning data and optimizing the learning path, specifically:

[0093] S1.1. Establish the loss function L(g(t),t) based on the user’s learning goals and actual eloquence performance.

[0094] According to the gap between the user's target score and the actual evaluation score, a basic cost function is established, and by introducing a linear penalty term into the basic cost function, an end cost function h(g(T)) is established;

[0095] Establish the target optimization model F according to the loss function L(g(t),t) and the terminal cost function h(g(T));

[0096] Setting learning goals through the goal optimization model F and establishing preset courses based on the learning goals; where the learning goals can be short-term (such as a training session) or long-term (such as a training cycle);

[0097] Among them, the target optimization model is:

[0098]

[0099] Among them, g(t) represents the learning goal at time t; L(g(t), t) is the loss function, which reflects the difference between the current learning goal and the user's ability; h(g(T)) is the terminal cost function, which represents the cost of not achieving the goal at the end of the training cycle; T is the set training cycle duration.

[0100] This embodiment S1.1 quantifies the gap between the user's expected target score and the evaluation score actually achieved during the training process through the terminal cost function, so that the user can clearly understand what kind of cost will be faced if the preset goal is not achieved at the end of the training, which helps the user to always maintain a clear understanding of the goal during the training process; if the value of the cost function is high, it means that the gap between the training result and the goal is large, and the training method or strategy needs to be further adjusted. Combining the loss function and the terminal cost function to establish a target optimization model to construct a preset course can provide clear guidance for the course design of eloquence training, determine which course elements or teaching methods are most effective in improving the user's eloquence skills, and thus design training courses in a targeted manner.

[0101] S1.2. Through dynamic system model Track and update the user's completion of the goals set in the preset course to obtain learning data; among them, the dynamic system model Allows the system to monitor learning progress in detail and adjust the learning path based on the user's real-time performance;

[0102] By integrating advanced mathematical models and algorithms to foresee the challenges that users may face, and adjusting the learning plan in advance with topological continuity and feedback, the learning path can be optimized;

[0103] The dynamic system model is:

[0104]

[0105] The process of topological continuity and feedback adjustment can be expressed as:

[0106] ψ:X→Y, continuous Opening ψ -1 (V) is also an open set in X

[0107] Among them, P(x, t) represents the learning progress at position x and time t; D is the diffusion coefficient, which reflects the propagation speed of learning progress; μ is the decay coefficient, which indicates the speed of forgetting or progress regression; f(x, t) is the external learning input, such as new learning materials or additional tutoring; V represents the open set, that is, every possible state set;

[0108] X and Y represent the topological space of learning objectives and the topological space of actual achieved states, respectively; ψ is a continuous mapping of the learning path, which is used to ensure the smoothness and adaptability of learning adjustments.

[0109] S2. Calculate the initial comprehensive incentive based on the learning progress of the learning data.

[0110] Step S2 of the embodiment of the present application is specifically as follows:

[0111] According to the behavior pattern and its corresponding time node, the behavior type of the user is identified from the learning data through the behavior pattern recognition model P1(x,t); the behavior pattern recognition model P1(x,t) is mainly used to monitor and identify the key behaviors of users on the platform in real time, such as completing courses, participating in discussions, etc.

[0112] Based on the behavior type and reward distribution mechanism, the cumulative reward amount R(t) is calculated according to the additional rewards obtained by the user for completing the preset learning tasks; wherein the reward distribution mechanism is used to automatically calculate and distribute rewards, such as points, badges, etc., according to the user's behavior and learning progress;

[0113] The initial comprehensive incentive is obtained by integrating the cumulative reward amount R(t) and the behavior type.

[0114] Among them, the behavior pattern recognition model is:

[0115]

[0116] The cumulative reward amount is:

[0117]

[0118] The initial comprehensive incentive is:

[0119]

[0120] Among them, P1(x,t) obtained by the behavior pattern recognition model represents the recognition probability under time t and behavior type x; λ i is the attenuation factor, reflecting the degree of reduction in the impact from the occurrence of the behavior to the current time; t i and x i Respectively represent the time and type of the i-th behavior; 1 {x＝xi} is the indicator function, when x is equal to x i The value is 1 when it is, otherwise it is 0;

[0121] R(t) is the cumulative reward at time t, κ is the weight factor of the behavior recognition score, and F j (t) is the additional reward function obtained by the user after completing a specific learning task j; ω j is the weight associated with task j, which shows the importance of rewards for different tasks;

[0122] α and β are the coefficients for adjusting static rewards and the coefficients for adjusting the impact of dynamic behavior changes, respectively. t is time, and x is a single behavior type.

[0123] This embodiment S2 analyzes the timestamp and type of user activities through a behavioral pattern recognition model, intelligently identifies the user's learning behavior type, helps understand the user's learning habits, preferences and progress, and provides a data basis for subsequent reward calculations. The calculation of the cumulative reward amount combines the user's real-time behavior and learning outcomes to ensure that the reward distribution is both fair and motivating. The initial comprehensive incentive is generated by integrating the calculation of the cumulative reward amount and the behavior type, which not only takes into account the user's effort, but also takes into account the user's learning behavior type and learning style, so the initial comprehensive incentive can well meet the user's needs and expectations.

[0124] S3. Obtain user feedback data on the initial comprehensive incentive, and adjust the incentive intensity of the initial comprehensive incentive according to the feedback data to obtain a comprehensive incentive.

[0125] Step S3 of the embodiment of the present application is specifically as follows:

[0126] Collect model F through feedback k Obtain user feedback data on comprehensive incentives; the feedback data can be either quantitative ratings or qualitative comments;

[0127] According to the change in user behavior and the average feedback score in the feedback data, the incentive intensity of the initial comprehensive incentive at the preset time is adjusted through the incentive adjustment dynamic model S(t) to obtain a single-time incentive;

[0128] The sensitivity of the initial comprehensive excitation from the initial moment to the preset moment is adjusted through the adaptive feedback loop mechanism Φ(t) to obtain the comprehensive moment excitation;

[0129] The comprehensive incentive is composed of single moment incentive and comprehensive moment incentive;

[0130] Among them, the feedback collection model is:

[0131]

[0132] The dynamic model of incentive adjustment is:

[0133]

[0134] The adaptive feedback loop mechanism is:

[0135] Φ(t)=∫0 t σ(s)·S(s)ds

[0136] Among them, F obtained by the feedback collection model k , represents the total feedback collected at the kth time point; feedback i (t k ) is the i-th user at time t kFeedback provided, N is the number of users who provided feedback;

[0137] The S(t) obtained by the incentive adjustment dynamic model is based on the feedback score at the current time t The incentive intensity is dynamically adjusted according to the change in user behavior ΔP(t); θ(t) is the adjustment factor, which is a parameter that mainly depends on time changes; is the average feedback score, which reflects the user's satisfaction with the current incentive mechanism; ΔP(t) is the amount of user behavior change, which indicates the change in behavior recognition data; λ and μ are weight parameters used to balance the impact of feedback score and behavior change on incentive adjustment;

[0138] Φ(t) obtained by the adaptive feedback loop mechanism represents the cumulative incentive adjustment effect; σ(s) is the sensitivity function of the incentive adjustment effect, which represents the sensitivity of the incentive adjustment at different time points; S(s) represents the incentive intensity calculated within a time period with s as a variable, and S(s) may be different for each moment s. The intensity is adjusted dynamically based on real-time data and feedback.

[0139] The incentive adjustment dynamic model in S3 of this embodiment can automatically adjust the incentive intensity by real-time monitoring and integrating user feedback and behavior data to ensure that the incentive strategy always meets the needs and feedback of users;

[0140] Moreover, dynamically adjusting the intensity of the initial comprehensive incentive at a specific preset moment according to the change in user behavior and the average feedback score can make the incentive more in line with the user's current needs and status. In addition, the single moment incentive, as a change in the intensity of the incentive at a specific moment, can achieve real-time adjustment of the incentive strategy. Adjusting the sensitivity of the initial comprehensive incentive from the initial moment to the preset moment means that it is possible to effectively pay attention to the trend and speed of the incentive change over time, ensure the continuity and consistency of the incentive plan in time, and avoid the adverse effects of sudden changes in incentives on users; and the comprehensive moment incentive as an adjustment method for the overall incentive effect can ensure the adaptability and sustainability of the incentive mechanism. This method allows the incentive system to respond not only to immediate feedback, but also to evaluate the long-term adjustment effect. It can more accurately adjust the incentive intensity according to the user's behavior and feedback data, and optimize the user's learning experience and motivation.

[0141] S4. Generate a first improvement suggestion based on the learning data and comprehensive incentives.

[0142] Step S4 of the embodiment of the present application is specifically as follows:

[0143] According to the learning data, the learning pattern recognition model M i (t) Identify users’ learning patterns, achievements and potential areas for improvement and obtain analytical results;

[0144] Based on the analysis results, use the report generation algorithm R to generate detailed progress reports and incentive effect reports;

[0145] Based on progress reports and incentive effect reports, the incentive mechanism is optimized and adjusted through a dynamic feedback loop mechanism D(t) to adapt to the actual needs and behavior changes of users;

[0146] Based on the adjusted and optimized learning path and incentive mechanism, the first improvement suggestion is generated according to the learning data and comprehensive incentives;

[0147] Among them, the learning pattern recognition model is:

[0148]

[0149] The report generation algorithm is:

[0150]

[0151] The dynamic feedback loop mechanism is:

[0152]

[0153] Among them, M obtained by learning the pattern recognition model i (t), represents the learning mode index of user i at time t; X i (s) is the activity data of user i at time s; i is the decay coefficient, reflecting the impact of past behavior on the current learning pattern; γ i is a regularization parameter used to adjust the weights of different user behavior data;

[0154] R obtained by the report generation algorithm is the generated comprehensive report indicator; M j (t) is the learning pattern analysis result of the jth user, ω j is the weight of the analysis result of the jth user; F k is the kth feedback indicator, such as user satisfaction or learning effect; η k is the weight of the kth feedback indicator;

[0155] D(t) obtained by the dynamic feedback loop mechanism represents the dynamic feedback of educational intervention at time t; σ(s) is the feedback sensitivity at time s; is the expected reporting benchmark, such as industry average or historical best performance; R(s) and R both refer to comprehensive reporting indicators, but R(s) represents the real-time comprehensive reporting indicator generated at different time points s, reflecting the integrated analysis results of user learning status and feedback effects at that time point, while R is an overall comprehensive reporting indicator.

[0156] The learning pattern recognition model in S4 of this embodiment helps to identify the user's learning habits and behavioral trends by analyzing time series data, so as to make personalized learning recommendations and interventions; and by segmenting the comprehensive report indicators by time, it can perform real-time effect evaluation and necessary adjustments.

[0157] S5. Establish a set of improvement suggestions based on the user behavior patterns and change trends in the learning data.

[0158] Step S5 of the embodiment of the present application includes S5.1 to S5.4; wherein S5.1 is a process of calculating the comprehensive data analysis score, S5.2 is a process of calculating the comprehensive analysis result, S5.3 is a process of establishing an improvement suggestion set based on the comprehensive data analysis score and the comprehensive analysis result, and S5.4 is a process of optimizing the improvement suggestion set, specifically:

[0159] S5.1. Capture the skills of the learning data through the preset eloquence performance analysis model B(t) to obtain the comprehensive score of eloquence performance;

[0160] According to the interaction methods and interaction times of users in the learning data, the interaction index is calculated through the user interaction data integration model I(t);

[0161] By integrating the comprehensive score of eloquence performance and the interaction index, the comprehensive data analysis score D(t) was obtained;

[0162] Among them, the eloquence performance analysis model is:

[0163]

[0164] The user interaction data integration model is:

[0165]

[0166] The comprehensive data analysis scores are:

[0167] D(t)=θ·B(t)+(1-θ)·I(t)

[0168] Among them, B(t) obtained by the eloquence performance analysis model represents the comprehensive score of eloquence performance at time t; x k (s) is the real-time measurement of the kth eloquence feature (such as speaking speed, pause) at time s; α k is the attenuation factor, which is used to adjust the timeliness of each feature; β k is the weight, which is used to adjust the influence of each feature; e is a natural constant, t is the total time, and t0 is the real time;

[0169] I(t) obtained by the user interaction data integration model represents the user interaction index at time t; c j(t) is the participation count of user in interaction j (e.g., posting, commenting), and n is the total number of interactions; γ j is the weight factor of interaction type j, which is used to characterize its importance in user evaluation;

[0170] θ is a preset parameter used to balance the importance of the two data sources.

[0171] In this embodiment S5.1, the eloquence performance analysis model allows the system to analyze and evaluate the user's eloquence performance in real time, providing the user with instant feedback and improvement suggestions; the user interaction data integration model can help the system identify the patterns and dynamics of user participation by quantifying different types of user interactions, thereby optimizing the platform content and interaction strategies.

[0172] S5.2. Perform preprocessing such as cleaning and standardization on the learning data to obtain the first data P(x i );

[0173] Use statistical analysis and machine learning techniques to extract valuable information and patterns from learning data and obtain prediction output f(x; Θ);

[0174] For the first data P(x i ) and the predicted output f(x; Θ) are used for data fusion and multi-dimensional analysis to obtain the comprehensive analysis result Z;

[0175] Among them, the data cleaning and standardization formula is:

[0176]

[0177] The machine learning model formula used by machine learning technology is:

[0178]

[0179] The formula for data fusion and multi-dimensional analysis is:

[0180] Z=ω1·P(x1)+ω2·f(x2;Θ)

[0181] Among them, x i represents the original data point; μ and σ are the mean and standard deviation of the learning data, respectively, which are used to standardize the data;

[0182] is the predicted output, x is the input feature vector; f is the preset machine learning model, decision tree; Θ represents the parameters of the machine learning model;

[0183] P(x1) and f(x2; Θ) represent the preprocessed data and the output of the machine learning model, respectively; ω1 and ω2 are weight parameters used to adjust the influence of different data sources in the analysis.

[0184] This embodiment S5.2 ensures the accuracy and reliability of the analysis results by removing noise and outliers in the data before analysis; the machine learning model allows the system to automatically learn and identify complex patterns and trends from large amounts of data, which can improve teaching methods and personalized learning experience.

[0185] S5.3, extract information from the comprehensive data analysis score D(t) and the comprehensive analysis result Z, and obtain an insight score I including learning weaknesses and strengths by adjusting the offset of the output information;

[0186] Based on the insight score I, the initial set of improvement suggestions S is obtained by calculating the probability of different given suggestions;

[0187] The overall feedback score R is calculated based on the insight score I and the weight of each suggestion in the initial improvement suggestion set S;

[0188] According to the overall feedback score R, several suggestions that meet the preset conditions are selected from the initial improvement suggestion set S to obtain an improvement suggestion set;

[0189] The insight scores are:

[0190]

[0191] The initial set of improvement suggestions is:

[0192]

[0193] The overall feedback score is:

[0194]

[0195] Among them, x j is the characteristic of user behavior data, w j is the feature weight, b is the bias term; σ is the activation function used to convert linear output into nonlinear output;

[0196] p(s|I) is the conditional probability of suggesting s given the insight;

[0197] λ i is the weight of the ith suggestion, m is the total number of suggestions in the initial set of improvement suggestions, and f i is a function that evaluates the performance of the i-th suggestion.

[0198] In the process of obtaining the insight score in S5.3 of this embodiment, the user behavior data is integrated by weighted sum, and the expression ability of the model is enhanced by using nonlinear transformation, which allows the system to have a deeper understanding of the dynamics behind the user behavior; by extracting information from the comprehensive data analysis score and the comprehensive analysis results, key learning performance data can be captured, which helps to have a deeper understanding of the user's learning status and discover potential improvement points, and then customize a highly targeted and practical initial improvement suggestion set based on the user's specific learning situation and potential problems, which fully considers the user's individual differences and specific needs;

[0199] The concept of an overall feedback score is also introduced. This score combines the insight score and the weight of each suggestion in the initial set of improvement suggestions. This helps to more comprehensively understand the effectiveness and feasibility of the initial set of suggestions, and to filter the initial set of improvement suggestions. This means that suggestions that better meet the actual needs of users can be dynamically selected.

[0200] S5.4. Generate progress-performance analysis report based on learning data g ;

[0201] Continuously monitor and evaluate the effectiveness of improvement measures for teaching content and learning experience, and obtain the effectiveness of improvement measures M f ;

[0202] According to the progress-performance analysis report R g and the effect of improvement measures f , the total effect O of teaching optimization is obtained by overall calculation; and the suggestion set is continuously optimized and improved according to the total effect O of teaching optimization;

[0203] Among them, the progress-performance analysis report is:

[0204]

[0205] The effects of the improvement measures are:

[0206]

[0207] The overall effect of teaching optimization is:

[0208]

[0209] Among them, D k is the k-th type of data after the learning data has been preprocessed by cleaning and standardization, α k is the weight of the k-th category of data in the report; K is the total number of data categories, and Visualize is a function used to convert data into charts and visualization elements;

[0210] S(t) is the learning satisfaction or effect at time t, β and γ are parameters for adjusting monitoring sensitivity, and T is the period of monitored time;

[0211] θ represents the set of adjustable teaching parameters; χ is a function used to evaluate the comprehensive effect of given teaching parameters, monitor results and generate reports; Θ is the space of all possible teaching parameters.

[0212] This embodiment S5.4 generates a comprehensive and intuitive learning progress-performance analysis report by converting various types of data into visual forms and weighted summarizing them; uses calculus to evaluate the rate of change of learning satisfaction or effect and its cumulative impact, and can monitor the actual effect of improvement measures; provides instant and dynamic learning feedback and progress analysis through the overall effect of teaching optimization, thereby optimizing the set of improvement suggestions.

[0213] Overall, this embodiment S5 can provide a quantitative indicator for the user's eloquence performance by capturing the skills in the learning data to obtain a comprehensive score of eloquence performance, thereby clearly understanding the user's eloquence skill level; the interaction index is calculated based on the user's interaction method and interaction times in the learning data, reflecting the user's participation and enthusiasm in the learning process, so the comprehensive data analysis score obtained by fusion on this basis can provide a more comprehensive evaluation of the user's learning effect. According to the comprehensive score of eloquence performance and the comprehensive analysis results, a set of improvement suggestions is established. These suggestions can put forward specific improvement suggestions for the user's eloquence skills and interaction methods, etc., to help users improve their eloquence ability more effectively.

[0214] S6. Generate an eloquence training program based on the first improvement suggestion and the improvement suggestion set.

[0215] Step S6 of the embodiment of the present application is specifically as follows:

[0216] An eloquence training program is generated from the first improvement suggestion and the improvement suggestion set, and eloquence training is performed on the user according to the eloquence training program.

[0217] Overall, this embodiment has the following beneficial effects:

[0218] The present invention establishes preset courses based on the terminal cost function, which can ensure that the course design meets the preset quality standards and realize the optimization of the course design. Feedback data provides the system with a basis for improving the incentive strategy. By analyzing the user's feedback, it is possible to identify which incentive measures are effective. Therefore, by adjusting the incentive intensity of the initial comprehensive incentive, it is possible to ensure that the incentive provided matches the user's expectations and needs. Since the learning data reflects the user's strengths, weaknesses and existing problems in eloquence learning, the comprehensive incentive after adjusting the intensity not only considers the user's actual performance, but also considers the user's psychological state and learning motivation. Therefore, the first eloquence improvement suggestion generated on this basis can stimulate the user's learning interest and motivation, so that the user can participate in eloquence learning more actively. By analyzing the user's behavior pattern and change trend in the learning data, the user's learning habits, points of interest and possible difficulties can be deeply understood. Based on these generated improvement suggestion sets, the user can be helped to locate the problem more accurately and provide effective solutions. Therefore, the eloquence training program generated by combining the first improvement suggestion and the improvement suggestion set not only makes a dynamic adjustment to the eloquence incentive machine, but also fully considers the user's eloquence strengths and weaknesses, thereby ensuring the effectiveness of the training effect;

[0219] In addition, by using deep learning and data analysis technology to extract useful insights from large amounts of user behavior data, it is possible to provide users with personalized learning suggestions and improvement strategies; by providing real-time, dynamic learning feedback and progress analysis through result display and reporting tools, it can help users and coaches better understand learning progress and optimize teaching strategies.

[0220] Embodiment 2:

[0221] See also Figure 2 , the embodiment of the present application provides an eloquence training device based on an incentive mechanism, including a data module 10, an incentive module 20, a feedback module 30, a suggestion module 40, a collection module 50 and a solution module 60;

[0222] The data module 10 is used to obtain the user's learning data on the preset course; wherein the preset course is established based on the user's learning goal and according to the terminal cost function;

[0223] An incentive module 20, configured to calculate an initial comprehensive incentive according to a learning progress of the learning data;

[0224] A feedback module 30 is used to obtain user feedback data on the initial comprehensive incentive, and adjust the incentive intensity of the initial comprehensive incentive according to the feedback data to obtain a comprehensive incentive;

[0225] A suggestion module 40, for generating a first improvement suggestion according to the learning data and the comprehensive incentive;

[0226] A collection module 50, for establishing a collection of improvement suggestions based on the behavior patterns and change trends of users in the learning data;

[0227] The solution module 60 is used to generate an eloquence training solution based on the first improvement suggestion and the improvement suggestion set.

[0228] In one embodiment, the data module 10 includes a loss unit, a gap unit, a course unit and a path unit; wherein the loss unit, the gap unit and the course unit are the process of establishing a preset course, and the path unit is the process of obtaining learning data and optimizing the learning path, specifically:

[0229] The loss unit is used to establish the loss function L(g(t),t) according to the user's learning goal and actual eloquence performance;

[0230] The gap unit is used to establish a basic cost function according to the gap between the user's target score and the actual evaluation score, and to establish a terminal cost function h(g(T)) by introducing a linear penalty term into the basic cost function;

[0231] The course unit is used to establish the target optimization model F based on the loss function L(g(t),t) and the terminal cost function h(g(T));

[0232] The course unit is also used to set learning goals through the goal optimization model F and establish preset courses according to the learning goals; wherein the learning goals can be short-term (such as a training meeting) or long-term (such as a training cycle);

[0233] Among them, the target optimization model is:

[0234]

[0235] Among them, g(t) represents the learning goal at time t; L(g(t), t) is the loss function, which reflects the difference between the current learning goal and the user's ability; h(g(T)) is the terminal cost function, which represents the cost of not achieving the goal at the end of the training cycle; T is the set training cycle duration.

[0236] The loss unit, gap unit and course unit of this embodiment quantify the gap between the user's expected target score and the evaluation score actually achieved during the training process through the terminal cost function, so that the user can clearly understand what kind of cost will be faced if the preset goal is not achieved at the end of the training, which helps the user to always maintain a clear understanding of the goal during the training process; if the value of the cost function is high, it means that the gap between the training result and the goal is large, and the training method or strategy needs to be further adjusted. Combining the loss function and the terminal cost function to establish a target optimization model to construct a preset course can provide clear guidance for the course design of eloquence training, determine which course elements or teaching methods are most effective in improving the user's eloquence skills, and thus design training courses in a targeted manner.

[0237] Path units are used to model dynamic systems Track and update the user's completion of the goals set in the preset course to obtain learning data; among them, the dynamic system model Allows the system to monitor learning progress in detail and adjust the learning path based on the user's real-time performance;

[0238] The path unit is also used to optimize the learning path by integrating advanced mathematical models and algorithms to foresee the challenges that users may face and make topological continuity and feedback adjustments to the learning plan in advance;

[0239] The dynamic system model is:

[0240]

[0241] The process of topological continuity and feedback adjustment can be expressed as:

[0242] ψ:X→Y, continuous Opening ψ -1 (V) is also an open set in X

[0243] Among them, P(x, t) represents the learning progress at position x and time t; D is the diffusion coefficient, which reflects the propagation speed of learning progress; μ is the decay coefficient, which indicates the speed of forgetting or progress regression; f(x, t) is the external learning input, such as new learning materials or additional tutoring; V represents the open set, that is, every possible state set;

[0244] X and Y represent the topological space of learning objectives and the topological space of actual achieved states, respectively; χ is a continuous mapping of the learning path, which is used to ensure the smoothness and adaptability of learning adjustments.

[0245] In one embodiment, the incentive module 20 includes a behavior unit, a reward unit, and a fusion unit;

[0246] Among them, the behavior unit is used to identify the user's behavior type from the learning data through the behavior pattern recognition model P1(x,t) according to the behavior pattern and its corresponding time node; Among them, the behavior pattern recognition model P1(x,t) is mainly used to monitor and identify the user's key behaviors on the platform in real time, such as completing courses, participating in discussions, etc.;

[0247] The reward unit is used to calculate the cumulative reward amount R(t) based on the additional rewards obtained by the user for completing the preset learning tasks based on the behavior type and the reward distribution mechanism; wherein the reward distribution mechanism is used to automatically calculate and distribute rewards, such as points, badges, etc., based on the user's behavior and learning progress;

[0248] The fusion unit is used to obtain the initial comprehensive incentive by fusing the cumulative reward amount R(t) and the behavior type.

[0249] Among them, the behavior pattern recognition model is:

[0250]

[0251] The cumulative reward amount is:

[0252]

[0253] The initial comprehensive incentive is:

[0254]

[0255] Among them, P1(x,t) obtained by the behavior pattern recognition model represents the recognition probability under time t and behavior type x; λ i is the attenuation factor, reflecting the degree of reduction in the impact from the occurrence of the behavior to the current time; t i and x i Respectively represent the time and type of the i-th behavior; 1 {x＝xi} is the indicator function, when x is equal to x i The value is 1 when it is, otherwise it is 0;

[0256] R(t) is the cumulative reward at time t, k is the weight factor of the behavior recognition score, and F j (t) is the additional reward function obtained by the user after completing a specific learning task j; ω j is the weight associated with task j, which shows the importance of rewards for different tasks;

[0257] α and β are the coefficients for adjusting static rewards and the coefficients for adjusting the impact of dynamic behavior changes, respectively. t is time, and x is a single behavior type.

[0258] The incentive module 20 of this embodiment analyzes the timestamp and type of user activities through the behavior pattern recognition model, intelligently identifies the user's learning behavior type, helps to understand the user's learning habits, preferences and progress, and provides a data basis for subsequent reward calculations. The calculation of the cumulative reward amount combines the user's real-time behavior and learning outcomes to ensure that the reward distribution is both fair and motivating. The initial comprehensive incentive is generated by integrating the calculation of the cumulative reward amount and the behavior type, which not only takes into account the user's level of effort, but also takes into account the user's learning behavior type and learning style, so the initial comprehensive incentive can well meet the user's needs and expectations.

[0259] In one embodiment, the feedback module 30 includes a single unit, an integrated unit, and a constituent unit;

[0260] Among them, a single unit is used to collect the model F through feedback k Obtain user feedback data on comprehensive incentives; the feedback data can be either quantitative ratings or qualitative comments;

[0261] The single unit is also used to adjust the incentive intensity of the initial comprehensive incentive at a preset moment through the incentive adjustment dynamic model S(t) according to the change amount of user behavior in the feedback data and the average feedback score, so as to obtain the incentive at a single moment;

[0262] A synthesis unit, used for adjusting the sensitivity of the initial synthesis excitation from the initial moment to the preset moment through an adaptive feedback loop mechanism Φ(t) to obtain a synthesis moment excitation;

[0263] A composition unit, used for composing a comprehensive incentive from a single moment incentive and a comprehensive moment incentive;

[0264] Among them, the feedback collection model is:

[0265]

[0266] The dynamic model of incentive adjustment is:

[0267]

[0268] The adaptive feedback loop mechanism is:

[0269] Φ(t)=∫0 t σ(s)·S(s)ds

[0270] Among them, F obtained by the feedback collection model k , represents the total feedback collected at the kth time point; feedback i (t k ) is the i-th user at time t k Feedback provided, N is the number of users who provided feedback;

[0271] The S(t) obtained by the incentive adjustment dynamic model is based on the feedback score at the current time t The incentive intensity is dynamically adjusted according to the change in user behavior ΔP(t); θ(t) is the adjustment factor, which is a parameter that mainly depends on time changes; is the average feedback score, which reflects the user's satisfaction with the current incentive mechanism; ΔP(t) is the amount of user behavior change, which indicates the change in behavior recognition data; λ and μ are weight parameters used to balance the impact of feedback score and behavior change on incentive adjustment;

[0272] Φ(t) obtained by the adaptive feedback loop mechanism represents the cumulative incentive adjustment effect; σ(s) is the sensitivity function of the incentive adjustment effect, which represents the sensitivity of the incentive adjustment at different time points; S(s) represents the incentive intensity calculated within a time period with s as a variable, and S(s) may be different for each moment s. The intensity is adjusted dynamically based on real-time data and feedback.

[0273] The incentive adjustment dynamic model in the feedback module 30 of this embodiment automatically adjusts the incentive intensity by real-time monitoring and integrating user feedback and behavior data, which can ensure that the incentive strategy always meets the needs and feedback of users;

[0274] Moreover, dynamically adjusting the intensity of the initial comprehensive incentive at a specific preset moment according to the change in user behavior and the average feedback score can make the incentive more in line with the user's current needs and status. In addition, the single moment incentive, as a change in the intensity of the incentive at a specific moment, can achieve real-time adjustment of the incentive strategy. Adjusting the sensitivity of the initial comprehensive incentive from the initial moment to the preset moment means that it is possible to effectively pay attention to the trend and speed of the incentive change over time, ensure the continuity and consistency of the incentive plan in time, and avoid the adverse effects of sudden changes in incentives on users; and the comprehensive moment incentive as an adjustment method for the overall incentive effect can ensure the adaptability and sustainability of the incentive mechanism. This method allows the incentive system to respond not only to immediate feedback, but also to evaluate the long-term adjustment effect. It can more accurately adjust the incentive intensity according to the user's behavior and feedback data, and optimize the user's learning experience and motivation.

[0275] In one embodiment, the suggestion module 40 is specifically:

[0276] According to the learning data, the learning pattern recognition model M i (t) Identify users’ learning patterns, achievements and potential areas for improvement and obtain analytical results;

[0277] Based on the analysis results, use the report generation algorithm R to generate detailed progress reports and incentive effect reports;

[0278] Based on progress reports and incentive effect reports, the incentive mechanism is optimized and adjusted through a dynamic feedback loop mechanism D(t) to adapt to the actual needs and behavior changes of users;

[0279] Based on the adjusted and optimized learning path and incentive mechanism, the first improvement suggestion is generated according to the learning data and comprehensive incentives;

[0280] Among them, the learning pattern recognition model is:

[0281]

[0282] The report generation algorithm is:

[0283]

[0284] The dynamic feedback loop mechanism is:

[0285]

[0286] Among them, M obtained by learning the pattern recognition model i (t), represents the learning mode index of user i at time t; X i (s) is the activity data of user i at time s; i is the decay coefficient, reflecting the impact of past behavior on the current learning pattern; γ i is a regularization parameter used to adjust the weights of different user behavior data;

[0287] R obtained by the report generation algorithm is the generated comprehensive report indicator; M j (t) is the learning pattern analysis result of the jth user, ω j is the weight of the analysis result of the jth user; F k is the kth feedback indicator, such as user satisfaction or learning effect; η k is the weight of the kth feedback indicator;

[0288] D(t) obtained by the dynamic feedback loop mechanism represents the dynamic feedback of educational intervention at time t; σ(s) is the feedback sensitivity at time s; is the expected reporting benchmark, such as industry average or historical best performance; R(s) and R both refer to comprehensive reporting indicators, but R(s) represents the real-time comprehensive reporting indicator generated at different time points s, reflecting the integrated analysis results of user learning status and feedback effects at that time point, while R is an overall comprehensive reporting indicator.

[0289] The learning pattern recognition model in the proposed module 40 of this embodiment helps to identify the user's learning habits and behavioral trends by analyzing time series data, so as to make personalized learning recommendations and interventions; and by segmenting the comprehensive report indicators by time, it can perform real-time effect evaluation and necessary adjustments.

[0290] In one embodiment, the collection module 50 includes a performance unit, an interaction unit, an analysis unit, a change unit, an insight subunit, a probability subunit, a weight subunit, an improvement subunit and a report unit; wherein the performance unit, the interaction unit and the analysis unit are processes for calculating the comprehensive data analysis score, the change unit is a process for calculating the comprehensive analysis result, the insight subunit, the probability subunit, the weight subunit and the improvement subunit are processes for establishing an improvement suggestion set based on the comprehensive data analysis score and the comprehensive analysis result, and the report unit is a process for optimizing the improvement suggestion set, specifically:

[0291] The performance unit is used to capture the skills of the learning data through the preset eloquence performance analysis model B(t) to obtain a comprehensive score of eloquence performance;

[0292] An interaction unit, used to calculate the interaction index through the user interaction data integration model I(t) according to the interaction mode and interaction times of the users in the learning data;

[0293] An analysis unit, used to obtain a comprehensive data analysis score D(t) by integrating the comprehensive score of eloquence performance and the interaction index;

[0294] Among them, the eloquence performance analysis model is:

[0295]

[0296] The user interaction data integration model is:

[0297]

[0298] The comprehensive data analysis scores are:

[0299] D(t)=θ·B(t)+(1-θ)·I(t)

[0300] Among them, B(t) obtained by the eloquence performance analysis model represents the comprehensive score of eloquence performance at time t; x k (s) is the real-time measurement of the kth eloquence feature (such as speaking speed, pause) at time s; α k is the attenuation factor, which is used to adjust the timeliness of each feature; β k is the weight, which is used to adjust the influence of each feature; e is a natural constant, t is the total time, and t0 is the real time;

[0301] I(t) obtained by the user interaction data integration model represents the user interaction index at time t; c j (t) is the participation count of user in interaction j (e.g., posting, commenting), and n is the total number of interactions; γ j is the weight factor of interaction type j, which is used to characterize its importance in user evaluation;

[0302] θ is a preset parameter used to balance the importance of the two data sources.

[0303] In the present unit, interaction unit and analysis unit of this embodiment, the eloquence performance analysis model allows the system to analyze and evaluate the user's eloquence performance in real time, providing the user with instant feedback and improvement suggestions; the user interaction data integration model can help the system identify the patterns and dynamics of user participation by quantifying different types of user interactions, thereby optimizing the platform content and interaction strategies.

[0304] The change unit is used to perform preprocessing such as cleaning and standardization on the learning data to obtain the first data P(x i );

[0305] The variation unit is also used to extract valuable information and patterns from the learning data using statistical analysis and machine learning techniques to obtain the predicted output f(x; Θ);

[0306] The changing unit is also used to change the first data P(x i ) and the predicted output f(x; Θ) are used for data fusion and multi-dimensional analysis to obtain the comprehensive analysis result Z;

[0307] Among them, the data cleaning and standardization formula is:

[0308]

[0309] The machine learning model formula used by machine learning technology is:

[0310]

[0311] The formula for data fusion and multi-dimensional analysis is:

[0312] Z=ω1·P(x1)+ω2·f(x2;Θ)

[0313] Among them, x i represents the original data point; μ and σ are the mean and standard deviation of the learning data, respectively, which are used to standardize the data;

[0314] is the predicted output; x is the input feature vector; f is the preset machine learning model, decision tree; Θ represents the parameters of the machine learning model;

[0315] P(x1) and f(x2; Θ) represent the preprocessed data and the output of the machine learning model, respectively; ω1 and ω2 are weight parameters used to adjust the influence of different data sources in the analysis.

[0316] The variation unit of this embodiment ensures the accuracy and reliability of the analysis results by removing noise and outliers in the data before analysis; the machine learning model allows the system to automatically learn and identify complex patterns and trends from large amounts of data, which can improve teaching methods and personalized learning experience.

[0317] The insight subunit is used to extract information from the comprehensive data analysis score D(t) and the comprehensive analysis result Z, and obtain the insight score I including the learning weaknesses and strengths by adjusting the offset of the output information;

[0318] The probability subunit is used to obtain an initial set of improvement suggestions S based on the insight score I and the probability of different suggestions;

[0319] The weight subunit is used to calculate the overall feedback score R according to the insight score I and the weight of each suggestion in the initial improvement suggestion set S;

[0320] An improvement subunit is used to select a number of suggestions that meet preset conditions from the initial improvement suggestion set S according to the overall feedback score R to obtain an improvement suggestion set;

[0321] The insight scores are:

[0322]

[0323] The initial set of improvement suggestions is:

[0324]

[0325] The overall feedback score is:

[0326]

[0327] Among them, x j is the characteristic of user behavior data, w j is the feature weight, b is the bias term; σ is the activation function used to convert linear output into nonlinear output;

[0328] p(s|I) is the conditional probability of suggesting s given the insight;

[0329] λ i is the weight of the ith suggestion, m is the total number of suggestions in the initial set of improvement suggestions, and f i is a function that evaluates the performance of the i-th suggestion.

[0330] In the process of obtaining insight scores by the insight subunit, probability subunit, weight subunit and improvement subunit of this embodiment, the user behavior data is integrated by weighted sum, and the expression ability of the model is enhanced by using nonlinear transformation, which allows the system to have a deeper understanding of the dynamics behind the user behavior; by extracting information from the comprehensive data analysis score and the comprehensive analysis results, key learning performance data can be captured, which helps to have a deeper understanding of the user's learning status and discover potential improvement points, and then customize a highly targeted and practical initial improvement suggestion set based on the user's specific learning situation and potential problems, which fully considers the user's individual differences and specific needs;

[0331] The concept of an overall feedback score is also introduced. This score combines the insight score and the weight of each suggestion in the initial set of improvement suggestions. This helps to more comprehensively understand the effectiveness and feasibility of the initial set of suggestions, and to filter the initial set of improvement suggestions. This means that suggestions that better meet the actual needs of users can be dynamically selected.

[0332] Reporting unit for generating progress-performance analysis reports based on learning data g ;

[0333] The reporting unit is also used to continuously monitor and evaluate the effectiveness of improvement measures for teaching content and learning experience, and obtain the effectiveness of improvement measures M f ;

[0334] Reporting unit, also used to report R based on progress-performance analysis g and the effect of improvement measures f , the total effect O of teaching optimization is obtained by overall calculation; and the suggestion set is continuously optimized and improved according to the total effect O of teaching optimization;

[0335] Among them, the progress-performance analysis report is:

[0336]

[0337] The effects of the improvement measures are:

[0338]

[0339] The overall effect of teaching optimization is:

[0340]

[0341] Among them, D k is the k-th type of data after the learning data has been preprocessed by cleaning and standardization, α k is the weight of the k-th category of data in the report; K is the total number of data categories, and Visualize is a function used to convert data into charts and visualization elements;

[0342] S(t) is the learning satisfaction or effect at time t, β and γ are parameters for adjusting monitoring sensitivity, and T is the period of monitored time;

[0343] θ represents the set of adjustable teaching parameters; ψ is a function used to evaluate the comprehensive effect of given teaching parameters, monitor results and generate reports; Θ is the space of all possible teaching parameters.

[0344] The reporting unit of this embodiment generates a comprehensive and intuitive learning progress-performance analysis report by converting various types of data into visual forms and weighted summarizing them; uses calculus to evaluate the rate of change of learning satisfaction or effect and its cumulative impact, and can monitor the actual effect of improvement measures; provides instant and dynamic learning feedback and progress analysis through the overall effect of teaching optimization, thereby optimizing the set of improvement suggestions.

[0345] Overall, the set module 50 of this embodiment can provide a quantitative indicator for the user's eloquence performance by capturing the skills in the learning data to obtain the comprehensive score of eloquence performance, thereby clearly understanding the user's eloquence skill level; the interaction index is calculated based on the user's interaction method and interaction times in the learning data, reflecting the user's participation and enthusiasm in the learning process, so the comprehensive data analysis score obtained by fusion on this basis can provide a more comprehensive evaluation of the user's learning effect. According to the comprehensive score of eloquence performance and the comprehensive analysis results, a set of improvement suggestions is established. These suggestions can put forward specific improvement suggestions for the user's eloquence skills and interaction methods, etc., to help users improve their eloquence ability more effectively.

[0346] In one embodiment, the solution module 60 is specifically:

[0347] An eloquence training program is generated from the first improvement suggestion and the improvement suggestion set, and eloquence training is performed on the user according to the eloquence training program.

[0348] Overall, this embodiment has the following beneficial effects:

[0349] The present invention establishes preset courses based on the terminal cost function, which can ensure that the course design meets the preset quality standards and realize the optimization of the course design. Feedback data provides the system with a basis for improving the incentive strategy. By analyzing the user's feedback, it is possible to identify which incentive measures are effective. Therefore, by adjusting the incentive intensity of the initial comprehensive incentive, it is possible to ensure that the incentive provided matches the user's expectations and needs. Since the learning data reflects the user's strengths, weaknesses and existing problems in eloquence learning, the comprehensive incentive after adjusting the intensity not only considers the user's actual performance, but also considers the user's psychological state and learning motivation. Therefore, the first eloquence improvement suggestion generated on this basis can stimulate the user's learning interest and motivation, so that the user can participate in eloquence learning more actively. By analyzing the user's behavior pattern and change trend in the learning data, the user's learning habits, points of interest and possible difficulties can be deeply understood. Based on these generated improvement suggestion sets, the user can be helped to locate the problem more accurately and provide effective solutions. Therefore, the eloquence training program generated by combining the first improvement suggestion and the improvement suggestion set not only makes a dynamic adjustment to the eloquence incentive machine, but also fully considers the user's eloquence strengths and weaknesses, thereby ensuring the effectiveness of the training effect;

[0350] In addition, by using deep learning and data analysis technology to extract useful insights from large amounts of user behavior data, it is possible to provide users with personalized learning suggestions and improvement strategies; by providing real-time, dynamic learning feedback and progress analysis through result display and reporting tools, it can help users and coaches better understand learning progress and optimize teaching strategies.

[0351] Embodiment three:

[0352] The embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the eloquence training method based on the incentive mechanism;

[0353] Wherein, the eloquence training method based on the incentive mechanism, if implemented in the form of a software functional unit and used as an independent product, can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0354] The above are preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. An eloquence training method based on an incentive mechanism, characterized in that: include: Acquire the user's learning data of a preset course; wherein the preset course is established based on the user's learning goal and according to the terminal cost function; Calculate an initial comprehensive incentive based on the learning progress of the learning data; Acquiring feedback data of the user on the initial comprehensive incentive, and adjusting the incentive intensity of the initial comprehensive incentive according to the feedback data to obtain a comprehensive incentive; generating a first improvement suggestion according to the learning data and the comprehensive incentive; Establishing a set of improvement suggestions based on the behavior patterns and change trends of users in the learning data; An eloquence training program is generated based on the first improvement suggestion and the improvement suggestion set.

2. The method for training eloquence based on an incentive mechanism as claimed in claim 1, characterized in that: The preset course is established based on the user's learning goal and the terminal cost function, specifically: Establishing a loss function according to the user's learning goal and actual eloquence performance; Establishing a basic cost function according to the gap between the target score and the actual evaluation score of the user, and establishing the terminal cost function by introducing a linear penalty term into the basic cost function; The preset course is established according to the loss function and the terminal cost function.

3. The method for training eloquence based on an incentive mechanism as claimed in claim 1, characterized in that: The initial comprehensive incentive is calculated according to the learning progress of the learning data, specifically: According to the behavior pattern and its corresponding time node, identifying the user's behavior type from the learning data; Based on the behavior type, a cumulative reward amount is calculated according to the additional reward obtained by the user for completing the preset learning task; The initial comprehensive incentive is obtained by merging and calculating the accumulated reward amount and the behavior type.

4. The method for training eloquence based on an incentive mechanism as claimed in claim 3, characterized in that: The initial comprehensive incentive is specifically: Among them, R(t) is the cumulative reward amount, P1(x,t) is the recognition probability of the behavior type, α and β are the coefficients for adjusting the static reward and the coefficients for adjusting the impact of dynamic behavior changes, t is time, and x is a single behavior type.

5. The method for training eloquence based on an incentive mechanism as claimed in claim 1, characterized in that: The incentive intensity of the initial comprehensive incentive is adjusted according to the feedback data to obtain a comprehensive incentive, specifically: According to the change amount of user behavior in the feedback data and the average feedback score, adjusting the incentive intensity of the initial comprehensive incentive at a preset time to obtain a single-time incentive; Adjusting the sensitivity of the initial comprehensive stimulus from the initial moment to the preset moment to obtain a comprehensive moment stimulus; The comprehensive incentive is composed of the single moment incentive and the comprehensive moment incentive.

6. The method for training eloquence based on an incentive mechanism as claimed in claim 1, characterized in that: According to the behavior patterns and change trends of users in the learning data, a set of improvement suggestions is established, specifically: Capturing the skills of the learning data through a preset eloquence performance analysis model to obtain a comprehensive eloquence performance score; Calculate the interaction index according to the interaction mode and interaction times of the users in the learning data; By integrating the comprehensive score of eloquence performance and the interaction index, a comprehensive data analysis score is obtained; Identify behavioral patterns and change trends in the learning data to obtain comprehensive analysis results; The improvement suggestion set is established according to the comprehensive data analysis score and the comprehensive analysis result.

7. The method for training eloquence based on an incentive mechanism as claimed in claim 6, characterized in that: According to the comprehensive data analysis score and the comprehensive analysis result, the improvement suggestion set is established, specifically: Extracting information from the comprehensive data analysis score and the comprehensive analysis result, and obtaining an insight score by adjusting an offset of the output information; Based on the insight score, an initial set of improvement suggestions is obtained according to probability calculation of given different suggestions; Calculating an overall feedback score according to the insight score and the weight of each suggestion in the initial set of improvement suggestions; According to the overall feedback score, several suggestions that meet a preset condition are selected from the initial set of improvement suggestions to obtain the set of improvement suggestions.

8. The method for training eloquence based on an incentive mechanism as claimed in claim 6, characterized in that: The eloquence performance analysis model is specifically: Among them, x k (s) is the real-time measurement value of the kth eloquence feature at time s, α k and β k are the attenuation factor and weight respectively, e is a natural constant, t is the total time, and t0 is the real time.

9. An eloquence training device based on an incentive mechanism, characterized in that: It includes data module, incentive module, feedback module, suggestion module, collection module and solution module; The data module is used to obtain the user's learning data on the preset course; wherein the preset course is established based on the user's learning goal and according to the terminal cost function; The incentive module is used to calculate an initial comprehensive incentive according to the learning progress of the learning data; The feedback module is used to obtain the user's feedback data on the initial comprehensive incentive, and adjust the incentive intensity of the initial comprehensive incentive according to the feedback data to obtain a comprehensive incentive; The suggestion module is used to generate a first improvement suggestion based on the learning data and the comprehensive incentive; The collection module is used to establish an improvement suggestion set according to the user's behavior pattern and change trend in the learning data; The solution module is used to generate an eloquence training solution based on the first improvement suggestion and the improvement suggestion set.

10. The eloquence training device based on incentive mechanism as claimed in claim 9, characterized in that: The data module includes loss units, gap units and course units; The loss unit is used to establish a loss function according to the user's learning goal and actual eloquence performance; The gap unit is used to establish a basic cost function according to the gap between the target score and the actual evaluation score of the user, and to establish the terminal cost function by introducing a linear penalty term into the basic cost function; The course unit is used to establish the preset course according to the loss function and the terminal cost function.

11. The eloquence training device based on incentive mechanism as claimed in claim 9, characterized in that: The incentive module includes a behavior unit, a reward unit and a fusion unit; The behavior unit is used to identify the user's behavior type from the learning data according to the behavior pattern and its corresponding time node; The reward unit is used to calculate the cumulative reward amount based on the behavior type and the additional reward obtained by the user for completing the preset learning task; The fusion unit is used to obtain the initial comprehensive incentive by performing a fusion calculation on the accumulated reward amount and the behavior type.

12. The eloquence training device based on incentive mechanism as claimed in claim 11, characterized in that: The initial comprehensive incentive is specifically: Among them, R(t) is the cumulative reward amount, P1(x,t) is the recognition probability of the behavior type, α and β are the coefficients for adjusting the static reward and the coefficients for adjusting the impact of dynamic behavior changes, t is time, and x is a single behavior type.

13. The eloquence training device based on incentive mechanism as claimed in claim 9, characterized in that: The feedback module includes a single unit, a comprehensive unit and a constituent unit; The single unit is used to adjust the incentive intensity of the initial comprehensive incentive at a preset time according to the change amount of user behavior in the feedback data and the average feedback score to obtain a single-time incentive; The synthesis unit is used to adjust the sensitivity of the initial comprehensive excitation from the initial moment to the preset moment to obtain a comprehensive moment excitation; The constituent unit is used to constitute the comprehensive excitation from the single moment excitation and the comprehensive moment excitation.

14. The eloquence training device based on incentive mechanism as claimed in claim 9, characterized in that: The assembly module includes a performance unit, an interaction unit, an analysis unit, a change unit and an assembly unit; The performance unit is used to perform skill capture on the learning data through a preset eloquence performance analysis model to obtain a comprehensive score of eloquence performance; The interaction unit is used to calculate the interaction index according to the interaction mode and interaction times of the user in the learning data; The analysis unit is used to obtain a comprehensive data analysis score by integrating the comprehensive score of eloquence performance and the interaction index; The change unit is used to identify the behavior pattern and change trend in the learning data to obtain a comprehensive analysis result; The collection unit is used to establish the improvement suggestion set according to the comprehensive data analysis score and the comprehensive analysis result.

15. The eloquence training device based on incentive mechanism as claimed in claim 14, characterized in that: The collection unit includes the insight subunit, the probability subunit, the weight subunit and the improvement subunit; The insight subunit is used to extract information from the comprehensive data analysis score and the comprehensive analysis result, and obtain an insight score by adjusting the offset of the output information; The probability subunit is used to obtain an initial set of improvement suggestions based on the insight score and according to the probability calculation of different given suggestions; The weight subunit is used to calculate the overall feedback score according to the insight score and the weight of each suggestion in the initial improvement suggestion set; The improvement subunit is used to select several suggestions that meet preset conditions in the initial improvement suggestion set according to the overall feedback score to obtain the improvement suggestion set.

16. The eloquence training device based on incentive mechanism as claimed in claim 14, characterized in that: The eloquence performance analysis model is specifically: Among them, x k (s) is the real-time measurement value of the kth eloquence feature at time s, α k and β k are the attenuation factor and weight respectively, e is a natural constant, t is the total time, and t0 is the real time.

17. A storage medium, characterized in that: The storage medium stores a computer program, which is called and executed by a computer to implement any one of the motivational mechanism-based eloquence training methods of claims 1 to 8.

Images