Knowledge tracking method and system based on meta-cognition and answering behaviors of learner
By introducing learner metacognition and question-answering behavior modeling into the deep KT model, combining gated neural networks and long and short-term memory networks, the problem of existing models ignoring metacognition and behavioral characteristics when predicting learners' knowledge states is solved, and the accuracy and interpretability of predictions are improved.
Patent Information
- Application Number
- CN202411701256.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-26
- Publication Date
- 2025-03-18
AI Technical Summary
When predicting learners' knowledge state, the existing deep KT model ignores the learners' metacognition, guessing, and misbehavior behaviors in the process of answering questions, resulting in inaccurate characterization of knowledge state.
By modeling the learner's knowledge state, extracting metacognitive information and modeling the answering behavior, combining gated neural networks and long-term memory networks, we can capture the changes in the learner's emotions, interaction behaviors and knowledge state during the answering process.
It improves the accuracy and interpretability of learners' performance predictions, and enhances the prediction ability of knowledge tracking models by more accurately capturing the complex dynamic changes of learners in the answering process.
Smart Images
Figure CN119647671A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of computer data processing, and in particular to a knowledge tracking method and system based on learners' metacognition and question-answering behavior. Background Art
[0002] Personalized education in the intelligent era requires providing learners with personalized, intelligent, and sustainable educational services. As a key technology for achieving personalized education, knowledge tracking aims to evaluate and track learners' knowledge status based on their historical learning performance, thereby predicting learners' performance in future exercises. Knowledge tracking has significant advantages in tracking learners' learning progress and optimizing teachers' teaching strategies, making it widely used in learning path planning, learning resource recommendation, and machine teaching.
[0003] The knowledge tracing model based on deep learning has become a research hotspot in the field of knowledge tracing due to its superior predictive performance. In recent years, a series of advanced deep KT models have emerged. For example, DKT introduces a recurrent neural network (RNN) to capture the complex process of learners' knowledge status changing over time. On this basis, the DIMKT model further integrates the difficulty information of exercises and knowledge concepts to dynamically build learners' personalized cognition. The IEKT model focuses on the accurate modeling of learners' cognition and knowledge acquisition process. LPKT monitors learners' knowledge status by directly modeling their learning process. KT-Deeper uses a recurrent neural network to learn the temporal characteristics of learners' cognitive evolution.
[0004] However, existing deep KT models still have some limitations. The above models usually use knowledge status directly for prediction, ignoring the learner's metacognition when answering, thus limiting the model's predictive performance. Metacognition refers to a person's awareness and control over his or her own cognitive process, including reflection and regulation of his or her own thinking, learning, and problem-solving processes. Studies have shown that metacognitive information such as learners' reflection and regulation of their own answering process affects learners' answering performance. Secondly, the existing KT model ignores the guesswork and mistakes made by learners in the process of answering questions when modeling the learner's knowledge status, resulting in an inaccurate portrayal of the learner's knowledge status. Summary of the invention
[0005] The main purpose of the present invention is to provide a knowledge tracking method and system based on learners' metacognition and answering behavior to solve the above technical problems.
[0006] To achieve the above object, the present invention provides a knowledge tracking method based on learners' metacognition and answering behavior, comprising the steps of: S1, modeling the learner's knowledge state, including: S11, based on the learner’s question-answering behavior in the question-answering step, a fully connected layer is used for deep fusion to obtain the basic interaction embedding ;in, is the question answered by the learner at time t, For Dimension is the result of the learner's answer at time t, For the dimension, , is a trainable parameter, represents the concatenation operation, R represents the set of real numbers, Represents dimension; S12, the embedding layer learns the embedding representation of the four emotion vectors, which are then merged into a comprehensive emotion representation embedding through the integration layer ;in, , is a trainable parameter, They represent the embedding of four different emotions respectively; S13, learning the number of prompts through the embedding layer , attempts , Answering time The embedding representation is merged into the number of prompts through the integration layer , attempts , Answering time Combined with basic interaction embedding to get comprehensive interaction embedding ;in, , is a trainable parameter; S2, extracting learners’ metacognitive information, including: capturing the complex relationship between learners’ emotions and responses to questions through a gated neural network to extract learners’ metacognitive information ; S3, Modeling learner behavior, including: via long short-term memory networks and comprehensive interaction embedding Capture the correlation between learners’ answer features and question responses to model learners’ guessing behavior and erroneous behavior ; S4, predicting learners’ subsequent performance in answering questions, including: S41, enhancing the learner's knowledge state in step S1 through the learner's behavior in step S3, and obtaining the learner's actual knowledge state at time t ; S42, obtain the questions answered by the learner at time t+1 , the corresponding adaptive knowledge state ; S43, the adaptive knowledge state and the learner's metacognitive information at time t Concatenate and predict the learner's answer to the question at time t+1 through the fully connected layer answering performance.
[0007] Furthermore, the step S2 comprises: S21, based on the formula: , , , as well as , calculate the learner's metacognitive sensitivity at time t and metacognitive bias ; S22, Metacognitive sensitivity of learners at time t and metacognitive bias Get the learner's metacognitive information about answering questions at time t ; in, , , the confidence level vector of the learner's judgment on the knowledge point to which the question belongs at time t , the learner's confidence level vector for answering this question at time t ;statistics The number of times is recorded as ; The number of times is recorded as ; The number of times is recorded as ; The number of times is recorded as ; HR is the hit rate, FAR is the false alarm rate, is the inverse cumulative distribution function of the standard normal distribution, , .
[0008] Furthermore, step S3 includes embedding the comprehensive interaction Input into the LSTM model to obtain the learner's guessing and error behaviors when answering questions: , , , A long short-term memory network specifically designed to detect guesses and mistakes. Indicates learners' guessing behavior when answering questions. Indicates learners' mistakes when answering questions.
[0009] Furthermore, the step S41 includes: Calculate the adjusted knowledge state ; For current issues The relevant knowledge association vector, For the previous time The state of knowledge of , is a knowledge correlation matrix, representing the relationship between problems and concepts; m represents the number of problems, i represents the number of knowledge concepts, and the elements of the knowledge correlation matrix are binary codes. Contains knowledge concepts hour, =1, otherwise =0; Through the input gate and the forget gate, simulate the learner's forgetting behavior of mastering knowledge , is a trainable parameter; Through the input gate and the forget gate, simulate the process of learners absorbing knowledge , is a trainable parameter; According to the formula , and obtain the actual knowledge state of the learner at time t .
[0010] Further, the step S42 includes: according to the formula , get the question answered by the learner at time t+1 Corresponding adaptive knowledge state .
[0011] Further, the step S43 includes: according to the prediction model , predict the learner's answer to the question at time t+1 performance in answering questions; In the range of (0,1), it means the question is answered correctly. The probability of , is a trainable parameter.
[0012] Furthermore, through the objective function , all parameters and vectors of the prediction model are trained.
[0013] The present invention also provides a knowledge tracking system based on learners' metacognition and answering behavior, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the knowledge tracking method based on learners' metacognition and answering behavior as described in any one of the above items are implemented.
[0014] The present invention also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the steps of the knowledge tracking method based on learner metacognition and answering behavior as described in any one of the above items are implemented: The present invention utilizes technical methods such as gated neural networks and long short-term memory networks to systematically analyze learners' historical practice interactions, answering time, number of prompts, etc., and integrates learners' guesses and error behaviors into knowledge tracking modeling with knowledge status as the core. At the same time, it systematically models learners' metacognition when answering questions and combines it with knowledge status to improve the accuracy and explainability of learners' performance predictions. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] The accompanying drawings are part of the present invention and are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but do not constitute an improper limitation of the present invention. Obviously, the accompanying drawings described below are only some embodiments. For ordinary technicians in this field, they can also obtain their own learning drawings based on these drawings without creative work. In the accompanying drawings: Figure 1 Schematic diagram of the flow of a knowledge tracking method based on learners' metacognition and question-answering behavior in one embodiment of the present invention.
[0016] Figure 2 The figure is a schematic diagram of the hardware structure for running the knowledge tracking method based on learners' metacognition and answering behavior in one embodiment of the present invention.
[0017] Figure 3 FIG. 1 is a logic diagram of step S2 in an embodiment of the present invention.
[0018] Figure 4 FIG. 1 is a logic diagram of step S3 in an embodiment of the present invention.
[0019] Figure 5 FIG. 4 is a logic diagram of step S41 in an embodiment of the present invention.
[0020] Figure 6 It is an overall logic diagram of a knowledge tracking method based on learners' metacognition and question-answering behavior in one embodiment of the present invention.
[0021] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0022] The following is a clear and complete description of the technical problems solved by the embodiments of the present invention, the technical solutions adopted, and the technical effects achieved in conjunction with the accompanying drawings and specific embodiments. 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 equivalent or obviously modified embodiments obtained by ordinary technicians in this field without creative work fall within the scope of protection of the present invention. The embodiments of the present invention can be concretized in a variety of different ways as defined and covered in the claims.
[0023] It should be noted that in the following description, many specific details are given for ease of understanding, but it is obvious that the present invention can be implemented without these specific details.
[0024] It should be noted that, in the absence of clear limitation or conflict, the various embodiments of the present invention and the technical features therein can be combined with each other to form a technical solution.
[0025] Please refer to Figures 1 to 6 ,The present invention proposes a knowledge tracking method based on learners' metacognition and answering behavior, including steps S1-S4.
[0026] S1, modeling the learner's knowledge state, including: S11, based on the learner’s question-answering behavior in the question-answering step, a fully connected layer is used for deep fusion to obtain the basic interaction embedding ;in, is the question answered by the learner at time t, For Dimension is the result of the learner's answer at time t, For the dimension, , is a trainable parameter, represents the concatenation operation, R represents the set of real numbers, Represents dimension; S12, the embedding layer learns the embedding representation of the four emotion vectors, which are then merged into a comprehensive emotion representation embedding through the integration layer ;in, , is a trainable parameter, They represent the embedding of four different emotions respectively; S13, learning the number of prompts through the embedding layer , attempts , Answering time The embedding representation is merged into the number of prompts through the integration layer , attempts , Answering time Combined with basic interaction embedding to get comprehensive interaction embedding ;in, , is a trainable parameter.
[0027] In step S11, , , Both represent dimensions, usually the dimensions of an embedding vector or feature vector, which are used to represent the dimensional size of the features processed by the model. R represents the set of real numbers, which is a basic mathematical symbol that represents a space that can take any real value. represents a real vector space of dimension dk, which means that it contains all vectors of dimension dk, and the elements of each vector are real numbers.
[0028] Please combine Figure 3 , is an emotion vector consisting of four true emotion representation values: concentration, boredom, confusion, and frustration. First, an embedding layer is applied to learn their respective embedding representations, which are represented as , , and Next, considering the different effects of these four emotions on the learning process and the differences in their influence, the embeddings of these four emotions are merged into a comprehensive emotion representation embedding through the integration layer. , to coordinate the diversity of various emotions. Those skilled in the art should know that the integration layer includes a series operation and a single-layer fully connected network, thereby ensuring a unified processing of different emotional factors.
[0029] In step 13, considering the different influences of different answer features on the learner’s answering process, an integration layer is used to merge the embeddings of these three answer features and basic interactions into a comprehensive interaction embedding , in order to effectively integrate this diversity.
[0030] S2, extracting learners’ metacognitive information, including: capturing the complex relationship between learners’ emotions and responses to questions through a gated neural network to extract learners’ metacognitive information .
[0031] Furthermore, the step S2 comprises: S21, based on the formula: , , , as well as , calculate the learner's metacognitive sensitivity at time t and metacognitive bias ; S22, Metacognitive sensitivity of learners at time t and metacognitive bias Get the learner's metacognitive information about answering questions at time t ; in, , , the confidence level vector of the learner's judgment on the knowledge point to which the question belongs at time t , the learner's confidence level vector for answering this question at time t ;statistics The number of times is recorded as ; The number of times is recorded as ; The number of times is recorded as ; The number of times is recorded as ; HR is the hit rate, FAR is the false alarm rate, is the inverse cumulative distribution function of the standard normal distribution, , .
[0032] Specifically, generally speaking, learners with higher metacognition will analyze questions carefully when answering them, constantly reflect on and adjust their methods during the answering process to improve their answering performance, while learners with lower metacognition may lack the ability to reflect and answer questions rigorously, resulting in poor answering performance. Therefore, quantifying metacognitive information can help the knowledge tracking model identify these differences and more accurately predict learners' answering performance.
[0033] Considering the complex interaction between learners’ emotions and responses to questions during a long learning process, and the excellent performance of the gated neural network GRU in capturing dependencies and changing trends in time series, GRU is used to extract the high-order relationship between learners’ historical responses to questions and emotions, that is, two confidence level vectors of learners’ answers at time t. Specifically, one is It represents the confidence level vector of the learner's judgment on the knowledge point of the question at time t, and the other is represents the learner's confidence level vector for answering this question at time t. These two confidence level vectors can reveal the complex dynamic changes of learners in the problem-solving process, for example, how learners' emotional fluctuations when encountering difficulties affect their problem-solving strategies and performance. The formula is as follows: in, , In order to facilitate the model to identify the confidence level, 0.5 is used as the threshold to distinguish high confidence from low confidence. There are four situations: High confidence question knowledge point judgment and high confidence correct answer , judge the knowledge point of the high-confidence question and answer it correctly with low confidence , judge the knowledge point of low-confidence questions and answer them correctly with high confidence , judge the knowledge point to which the low-confidence question belongs and answer the question correctly with low confidence . Count the number of four types of situations in these two vectors ( ), a confidence level cross table can be obtained, as shown in Table 1 below.
[0034] Table 1 Cross-tabulation of confidence levels Through this confidence level cross-tabulation, the hit rate (HR) (the proportion of learners who are confident that they can answer correctly when they are confident in their judgment of the knowledge point to which the question belongs) and the false alarm rate (FAR) (the proportion of learners who are not confident that they can answer correctly but are not confident when they are not confident in their judgment of the knowledge point to which the question belongs) can be calculated.
[0035] According to the hit rate and false alarm rate, metacognitive sensitivity and metacognitive bias are calculated: in, is the inverse cumulative distribution function of the standard normal distribution, represents the metacognitive sensitivity of the learner at time t, Represents the learner's metacognitive bias at time t.
[0036] Finally, by concatenating the obtained metacognitive sensitivity and metacognitive bias vectors, we can obtain the metacognitive information of the learner answering the question at time t: in, is a trainable parameter.
[0037] S3, modeling learner behavior, including: through long short-term memory networks (LSTM) and comprehensive interaction embedding Capture the correlation between learners’ answer features and question responses to model learners’ guessing behavior and erroneous behavior .
[0038] Furthermore, the step S3 includes embedding the comprehensive interaction into Input into the LSTM model to obtain the learner's guessing and error behaviors when answering questions: , , , A long short-term memory network specifically designed to detect guesses and mistakes. Indicates learners' guessing behavior when answering questions. Indicates learners' mistakes when answering questions.
[0039] Different characteristics of learners’ answers (number of attempts, number of prompts, answering time) work together to affect guessing and error behaviors. For example, if a learner answers a question very quickly, it does not necessarily mean that the learner is guessing. Perhaps the learner is just very proficient in the corresponding concept. However, if the learner also has other learner behaviors, such as repeated attempts, then the learner is more likely to guess a random answer through rapid attempts. Therefore, it is particularly important to capture the complex dependency patterns between different behaviors and simulate their synergistic effects on learners’ guessing and error behaviors.
[0040] LSTM is used to analyze the comprehensive interactive embedding containing answer features. Due to its unique memory unit and gating mechanism structure, LSTM can effectively capture and retain long-term dependency information. It has significant advantages in processing time series data and can effectively capture the complex dependency relationship of answer features in the sequence over time. Specifically, the comprehensive interactive embedding Input into the LSTM model to obtain the learner's guessing and error behaviors when answering questions. The step-by-step process is outlined as follows: in , A long short-term memory network specifically designed for detecting guesses and mistakes.
[0041] S4, predicting learners’ subsequent performance in answering questions, including: S41, enhancing the learner's knowledge state in step S1 through the learner's behavior in step S3, and obtaining the learner's actual knowledge state at time t ; S42, obtain the questions answered by the learner at time t+1 , the corresponding adaptive knowledge state ; S43, the adaptive knowledge state and the learner's metacognitive information at time t Concatenate and predict the learner's answer to the question at time t+1 through the fully connected layer answering performance.
[0042] Furthermore, the step S41 includes: Calculate the adjusted knowledge state ; For current issues The relevant knowledge association vector, For the previous time The state of knowledge of , is a knowledge correlation matrix, representing the relationship between problems and concepts; m represents the number of problems, i represents the number of knowledge concepts, and the elements of the knowledge correlation matrix are binary codes. Contains knowledge concepts hour, =1, otherwise =0; Through the input gate and the forget gate, simulate the learner's forgetting behavior of mastering knowledge , is a trainable parameter; Through the input gate and the forget gate, simulate the process of learners absorbing knowledge , is a trainable parameter; According to the formula , and obtain the actual knowledge state of the learner at time t .
[0043] Please combine the specific Figure 4 , each interaction of the learner usually involves different problems, resulting in different updates of the corresponding knowledge state. In order to match the interaction and knowledge state at time t, we first compare the knowledge association vector related to the current problem with the knowledge association vector at the previous time. The state of knowledge Multiply them together to generate the adjusted knowledge state : where · represents the corresponding element-wise product between vectors.
[0044] Inspired by the LSTM structure and LPKT model, the input gate and forget gate are used to simulate the learner's knowledge absorption and forgetting in the long-term learning stage. The input gate is used to capture the learner's acquisition of new knowledge, while the forget gate controls the gradual forgetting of the acquired knowledge, so as to be closer to the dynamic changes of the real learning process. Figure 4As shown in the figure, first, in the knowledge forgetting part, the learner’s current comprehensive interaction embedding, the adaptive knowledge state at the previous moment, and the learner’s answering time are combined, and then the obtained vector is input into the forgetting gate to simulate the learner’s forgetting behavior of mastering the knowledge: in, is a trainable parameter.
[0045] For the knowledge acquisition part, the learner's current comprehensive answering interaction, number of prompts, number of attempts, adaptive knowledge state at the previous moment, and response time are combined in the same way. Then, the process of simulating the learner's knowledge absorption is as follows: in, is a trainable parameter.
[0046] When learners answer questions correctly by guessing, they may not really understand the relevant knowledge points, resulting in an unstable knowledge structure. Similarly, when learners answer questions incorrectly due to mistakes, their understanding of the relevant knowledge points may be further deepened and consolidated. Therefore, considering the learners' guesses and mistakes, in order to obtain the learners' new knowledge state after interacting with the learning at time t ,we will , Respectively , Multiply them together to get the actual knowledge acquired by the learner, and add the influence of the forgetting factor to the knowledge state at the previous moment: Furthermore, in order to more accurately capture the learner's knowledge mastery when answering the question, firstly, the obtained knowledge state is multiplied by the question to be predicted at time t+1 to obtain the adaptive knowledge state corresponding to the question. The step S42 includes: according to the formula , get the question answered by the learner at time t+1 Corresponding adaptive knowledge state .
[0047] Furthermore, considering that different learners may show different reflections and adjustments in the same knowledge state due to different metacognition in the problem-solving process, the adaptive knowledge state and the learner's metacognitive information at time t Concatenate and predict the learner's answer to the question at time t+1 through the fully connected layer This combination reflects the learner's real ability and status, which helps the model to more accurately predict the learner's answering performance; the step S43 includes: , predict the learner's answer to the question at time t+1 performance in answering questions; In the range of (0,1), it means the question is answered correctly. The probability of , is a trainable parameter.
[0048] Furthermore, the cross entropy logarithmic loss between the predicted answer y and the actual answer r is used as the objective function. , all parameters and vectors of the prediction model are trained and minimized during the training process.
[0049] The present invention utilizes technical methods such as gated neural networks and long short-term memory networks to systematically analyze learners' historical practice interactions, answering time, number of prompts, etc., and integrates learners' guesses and error behaviors into knowledge tracking modeling with knowledge status as the core. At the same time, it systematically models learners' metacognition when answering questions and combines it with knowledge status to improve the accuracy and explainability of learners' performance predictions.
[0050] The present invention also provides a knowledge tracking system based on learners' metacognition and answering behavior, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the knowledge tracking method based on learners' metacognition and answering behavior as described in any one of the above items are implemented.
[0051] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the knowledge tracking method based on learner metacognition and answering behavior as described in any of the above items.
[0052] Figure 2 FIG. 1 is a schematic diagram of the hardware structure for running a knowledge tracking method based on learner metacognition and answering behavior provided by an embodiment of the present invention. Figure 2 As shown, the embodiment / computer 6 includes: a processor 60, a memory 61, and a computer program 62 stored in the memory 61 and executable on the processor 60, such as a program for executing a knowledge tracking method based on learner metacognition and answering behavior. When the processor 60 executes the computer program 62, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor 60 executes the computer program 62, the functions of the modules / units in the above-mentioned device embodiments are implemented.
[0053] Exemplarily, the computer program 62 may be divided into one or more modules / units, which are stored in the memory 61 and executed by the processor 60 to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program 62 in the computer 6.
[0054] The computer 6 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The computer 6 may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art will appreciate that Figure 2 It is only an example of computer 6 and does not constitute a limitation of computer 6. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer 6 may also include input and output devices, network access devices, buses, etc.
[0055] The processor 60 may be a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.
[0056] The memory 61 may be an internal storage unit of the computer 6, such as a hard disk or memory of the computer 6. The memory 61 may also be an external storage device of the terminal device / computer 6, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the terminal device. Further, the memory 61 may also include both an internal storage unit of the computer 6 and an external storage device. The memory 61 is used to store the computer program and other learner programs and data required by the terminal device. The memory 61 may also be used to temporarily store data that has been output or is to be output.
[0057] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.
[0058] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.
[0059] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0060] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminal equipment and methods can be implemented in other ways. For example, the device / terminal equipment embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.
[0061] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0062] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0063] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this 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. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, 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. 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. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.
[0064] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.
Claims
1. A knowledge tracking method based on learners' metacognition and answering behavior, characterized in that: Includes steps: S1, modeling the learner’s knowledge state, including: S11, based on the learner’s question-answering behavior in the question-answering step, a fully connected layer is used for deep fusion to obtain the basic interaction embedding ;in, is the question answered by the learner at time t, For the dimension, is the result of the learner's answer at time t, For the dimension, , is a trainable parameter, represents the connection operation, R represents the real number set, Represents dimension; S12, the embedding layer learns the embedding representation of the four emotion vectors, which are then merged into a comprehensive emotion representation embedding through the integration layer ;in, , is a trainable parameter, They represent the embedding of four different emotions respectively; S13, learning the number of prompts through the embedding layer , Number of attempts , Answering time The embedding representation is merged into the number of prompts through the integration layer , attempts , Answering time Combined with basic interaction embedding to get comprehensive interaction embedding ;in, , is a trainable parameter; S2, extracting learners’ metacognitive information, including: capturing the complex relationship between learners’ emotions and responses to questions through a gated neural network to extract learners’ metacognitive information ; S3, Modeling learner behavior, including: via long short-term memory networks and comprehensive interaction embedding Capture the correlation between learners’ answer features and question responses to model learners’ guessing behavior and erroneous behavior ; S4, predicting learners’ subsequent performance in answering questions, including: S41, enhancing the learner's knowledge state in step S1 through the learner's behavior in step S3, and obtaining the learner's actual knowledge state at time t ; S42, obtain the questions answered by the learner at time t+1 , the corresponding adaptive knowledge state ; S43, the adaptive knowledge state and the learner's metacognitive information at time t Concatenate and predict the learner's answer to the question at time t+1 through the fully connected layer answering performance.
2. The knowledge tracking method based on learner metacognition and answering behavior according to claim 1 is characterized in that: The step S2 comprises: S21, based on the formula: , , , as well as , calculate the learner's metacognitive sensitivity at time t and metacognitive bias ; S22, Metacognitive sensitivity of learners at time t and metacognitive bias Get the learner's metacognitive information about answering questions at time t ; in, , , the confidence level vector of the learner's judgment on the knowledge point to which the question belongs at time t , the learner's confidence level vector for answering this question at time t ;statistics The number of times is recorded as ; The number of times is recorded as ; The number of times is recorded as ; The number of times is recorded as ; HR is the hit rate, FAR is the false alarm rate, is the inverse cumulative distribution function of the standard normal distribution, , .
3. The knowledge tracking method based on learner metacognition and answering behavior according to claim 1 is characterized in that: The step S3 includes embedding the comprehensive interaction into Input into the LSTM model to obtain the learner's guessing and error behaviors when answering questions: , , , A long short-term memory network specifically designed to detect guesses and mistakes. Indicates learners' guessing behavior when answering questions. Indicates learners' mistakes when answering questions.
4. The knowledge tracking method based on learner metacognition and answering behavior according to claim 1 is characterized in that: The step S41 comprises: Calculate the adjusted knowledge state ; For current issues The relevant knowledge association vector, For the previous time The state of knowledge of , is a knowledge correlation matrix, representing the relationship between problems and concepts; m represents the number of problems, i represents the number of knowledge concepts, and the elements of the knowledge correlation matrix are binary codes. Contains knowledge concepts hour, =1, otherwise =0; Through the input gate and the forget gate, simulate the learner's forgetting behavior of mastering knowledge , is a trainable parameter; Through the input gate and the forget gate, simulate the process of learners absorbing knowledge , is a trainable parameter; According to the formula , and obtain the actual knowledge state of the learner at time t .
5. The knowledge tracking method based on learner metacognition and answering behavior according to claim 4 is characterized in that: The step S42 comprises: according to the formula , get the question answered by the learner at time t+1 Corresponding adaptive knowledge state .
6. The knowledge tracking method based on learner metacognition and answering behavior according to claim 5 is characterized in that: The step S43 includes: according to the prediction model , predict the learner's answer to the question at time t+1 performance in answering questions; In the range of (0,1), it means the question is answered correctly. The probability of , is a trainable parameter.
7. The knowledge tracking method based on learner metacognition and answering behavior according to claim 6 is characterized in that: Through the objective function , all parameters and vectors of the prediction model are trained.
8. A knowledge tracking system based on learners' metacognition and answering behavior, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the knowledge tracking method based on learner metacognition and question-answering behavior are implemented as described in any one of claims 1 to 7.
9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the knowledge tracking method based on learner metacognition and question-answering behavior are implemented as described in any one of claims 1 to 7.