Knowledge tracking method and system supporting dynamic context

Abstract

The invention belongs to the technical field of knowledge tracking, and discloses a knowledge tracking method and system supporting a dynamic context, and the method comprises the steps: constructing a knowledge tracking model supporting the length of the dynamic context based on a question set of a student; and based on the knowledge tracking model supporting the dynamic context length, obtaining a final student answering prediction result. Through a question encoder and an answer encoder, a historical answer sequence of a student is mapped into high-dimensional feature embedding, and a question level perception vector and joint question-answer interaction representation are generated. The attention score matrix is dynamically adjusted based on the directional penalty function matrix, and the adaptability of the model to the dynamic change of the context length is enhanced. A multi-head self-attention mechanism and zoom dot product attention are utilized to extract knowledge states of time sensitivity and context dependence from student interaction data.

Description

Technical Field

[0001] The present invention relates to the technical field of knowledge tracing, and in particular to a knowledge tracing method and system supporting dynamic context. Background Art

[0002] Knowledge tracing, as a core technology in personalized learning and intelligent education systems, aims to predict students' mastery of new knowledge points by analyzing their historical behaviors during the learning process. The goal of knowledge tracing is to dynamically capture changes in students' knowledge status and provide support for the design of personalized learning paths by accurately predicting students' mastery of knowledge points. Knowledge tracing can provide students with customized recommendations for content that needs to be strengthened to avoid repeating the knowledge they have already mastered. Knowledge tracing can help teachers monitor students' learning status in real time, identify learning difficulties in a timely manner, and adjust teaching strategies. By providing personalized support for students of different ability levels, knowledge tracing can also help narrow the learning gap and promote educational equity.

[0003] Although existing knowledge tracking models based on attention mechanisms (such as AKT, simpleKT, DTransformer, etc.) have made significant progress in performance, they all have a key problem: the model fixes the context length in both the training and prediction stages, and cannot effectively cope with the diversity of student learning path lengths in actual educational scenarios. For example, some students may have only dozens of learning interactions, while others may have accumulated thousands of interaction records. This limitation causes the model's performance to drop significantly when encountering longer interaction sequences than those in the training stage, restricting the possibility of using knowledge tracking models to achieve large-scale personalized education in actual educational scenarios. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a knowledge tracking method and system that supports dynamic context. Through the question encoder and the answer encoder, the student's historical answer sequence is mapped into a high-dimensional feature embedding to generate a question-level perception vector and a joint question-answer interaction representation. The attention score matrix is ​​dynamically adjusted based on the directional penalty function matrix to enhance the model's adaptability to dynamic changes in context length. The multi-head self-attention mechanism and scaled dot product attention are used to extract time-sensitive and context-dependent knowledge states from student interaction data.

[0005] The present invention discloses a knowledge tracking method supporting dynamic context, the method comprising:

[0006] Based on the students' question sets, a knowledge tracking model that supports dynamic context length is constructed;

[0007] Based on the knowledge tracking model supporting dynamic context length, the prediction result of the final student answer is obtained.

[0008] Preferably, the knowledge tracking model supporting dynamic context length includes: a student learning behavior representation module, a knowledge state extraction module, a directional weighted offset module, and a student answer prediction module;

[0009] The student learning behavior representation module is used to describe the student's learning behavior pattern in detail by constructing a multi-dimensional feature embedding space based on the representation learning theory;

[0010] The knowledge state extraction module is used to capture the deep-level student knowledge state based on the multi-head self-attention mechanism and dot product attention;

[0011] The directional weighted offset module is used to dynamically adjust the attention score matrix by introducing a directional weighted penalty function matrix to achieve dynamic adaptation of different context lengths;

[0012] The student answer prediction module is used to project the high-dimensional to low-dimensional knowledge state characteristics with a two-layer nonlinear fully connected network as the core, optimize the binary cross entropy loss function, dynamically adjust the model parameters, and obtain the final prediction result.

[0013] Preferably, based on representation learning theory, a multi-dimensional feature embedding space is constructed to finely describe the learning behavior patterns of students, including:

[0014] For student S, the historical learning behavior includes T interaction processes, each of which is expressed as t {c t}, r t >, where t∈T, specifically, q t represents the question answered by the student at time step t, c t represents the question q answered by the student at time step t t The corresponding knowledge point. If a question corresponds to multiple knowledge points, then it is the set of multiple knowledge points {c t}, r t Indicates that the student answers question q at time step t t The answer is r t =1 means the student answered correctly, r t =0 means the student answered incorrectly;

[0015] For the questions answered by students, the encoder E(·) is used to perform feature mapping on the question sequence to obtain a high-dimensional vector representation of question-level perception:

[0016]

[0017] in,​ is the level of questions that can be trained, is the average feature representation of all knowledge points corresponding to the question, is the unique hot encoding of the knowledge point corresponding to the question, ⊙ represents bitwise multiplication, Represents bitwise addition;

[0018] For the student’s answer, the encoder is used to model the high-order interaction characteristics between the question knowledge points and the student’s answer through the embedding layer, thereby generating a joint question-answer interaction representation:

[0019]

[0020] Among them, v(c t , r t ) represents the average feature representation of all knowledge points corresponding to the question of the joint student answers, e(c t , r t ) represents the unique hot encoding of the knowledge points corresponding to the joint students’ corresponding questions.

[0021] Preferably, the deep-level student knowledge state captured based on the multi-head self-attention mechanism and dot product attention includes:

[0022] Represent the high-dimensional vector x of question-level perception t and the joint question-answer interaction representation y t They are respectively used as query vector Q, key vector K and value vector V, namely:

[0023] Q = x t+1 ; K = {x1, ..., x t}; V = {y1, ..., y t};

[0024] A multi-head attention mechanism integrating scaled dot-product attention is used to extract the knowledge state h from the student's historical answer sequence. t+1 :

[0025]

[0026] h t+1 =Concat(Head1, Head2, .., Head n );

[0027] Among them, Head represents the multi-head representation in the multi-head attention mechanism, Softmax represents the softmax function, represents the transpose of the key vector, represents the scaling factor of the scaled dot product attention, and d is the dimension of the key vector.

[0028] Preferably, dynamically adjusting the attention score matrix by introducing a directional weighted penalty function matrix to achieve dynamic adaptation of different context lengths includes:

[0029] The attention score matrix is ​​calculated using the cosine similarity between the query vector and the key vector:

[0030]

[0031] The attention score matrix is ​​dynamically adjusted through the directional penalty function matrix F:

[0032]

[0033] Among them, f ij is the element of the directional penalty function matrix F, i and j are the indices of the attention score matrix elements, δ is the directional coefficient that can be obtained through training, and the positive and negative values ​​represent the positive and negative directions of the penalty. is a long representation with directionally weighted penalties, It is a knowledge state that has been directionally weighted penalized.

[0034] Preferably, with a two-layer nonlinear fully connected network as the core, the high-dimensional to low-dimensional projection of the knowledge state features is performed, the binary cross entropy loss function is optimized, and the model parameters are dynamically adjusted to obtain the final prediction results including:

[0035] With a two-layer nonlinear fully connected network as the core, the knowledge state is realized The high-dimensional to low-dimensional projection of Optimize and dynamically adjust model parameters to obtain the final prediction results of students’ answers

[0036]

[0037] Among them, σ represents the Sigmoid function, k1, k2, W1, W2 are learnable parameters, and r t+1 The actual answer results for the students.

[0038] The present invention also discloses a knowledge tracking system supporting dynamic context, the system is used to implement any one of the methods described above, the system comprising: a construction module and a prediction module;

[0039] The building module is used to build a knowledge tracking model supporting dynamic context length based on the student's question set;

[0040] The prediction module is used to obtain the prediction result of the final student answer based on the knowledge tracking model supporting dynamic context length.

[0041] The present invention also discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the methods described above when executing the program.

[0042] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed, any one of the methods described above is implemented.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The model of the present invention can dynamically adapt to the context lengths of students' different learning interaction sequences through a directional weighted penalty module combined with a directional penalty function matrix, significantly improving the application effect of the model in dynamic educational scenarios.

[0045] 2. The student learning behavior representation module uses the question encoder and the answer encoder to map the student's historical answer sequence into high-dimensional semantic features, and integrates the multi-dimensional information of questions, knowledge points and answer characteristics, providing accurate and efficient input representation.

[0046] 3. The knowledge state extraction module uses multi-head self-attention mechanism and scaled dot product attention to efficiently extract time sensitivity and context embedding characteristics in parallel computing, while capturing high-order feature interactions, providing deep support for student knowledge state modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solution of the present invention, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 A schematic diagram of a knowledge tracking method supporting dynamic context according to an embodiment of the present invention;

[0049] Figure 2 A schematic diagram of the structure of a knowledge tracking model supporting dynamic context length according to an embodiment of the present invention;

[0050] Figure 3 Schematic diagram of the structure of an electronic device according to an embodiment of the present invention.

[0052] 1010, processor; 1020, memory; 1030, input / output interface; 1040, communication interface; 1050, bus. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present disclosure should be understood by people with ordinary skills in the field to which the present disclosure belongs. The "first", "second" and similar words used in the embodiments of the present disclosure do not indicate any order, quantity or importance, but are only used to distinguish different components. "Including" or "comprising" and similar words mean that the elements or objects appearing before the word cover the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0055] First, some technical terms used in the present invention are explained:

[0056] Context length refers to the maximum length of text that the model can handle, that is, the maximum length of text that the model can understand and generate without affecting performance. In natural language processing (NLP), long context is crucial for the model to understand complex sentences and paragraphs, because it allows the model to consider more contextual information to make accurate predictions or generation.

[0057] The knowledge tracing model is an educational technology that aims to predict students' mastery and performance of future knowledge points by analyzing their learning history and behavior data. It is often used in intelligent teaching systems (ITS) to provide personalized learning experiences and adaptive teaching. The main types of knowledge tracing models: 1 Bayesian knowledge tracing (BKT), 2 Deep learning-based knowledge tracing (DKT), 3 Structure-aware inductive knowledge tracing (SINKT), 4 Stable knowledge tracing model (SKT-LT).

[0058] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0059] Embodiment 1

[0060] As can be seen from the background technology,

[0061] Although existing knowledge tracking models based on attention mechanisms (such as AKT, simpleKT, DTransformer, etc.) have made significant progress in performance, they all have a key problem: the model fixes the context length in both the training and prediction stages, and cannot effectively cope with the diversity of student learning path lengths in actual educational scenarios. For example, some students may have only dozens of learning interactions, while others may have accumulated thousands of interaction records. This limitation causes the model's performance to drop significantly when encountering longer interaction sequences than those in the training stage, restricting the possibility of using knowledge tracking models to achieve large-scale personalized education in actual educational scenarios.

[0062] like Figure 1 As shown, an embodiment of the present invention provides a knowledge tracking method supporting dynamic context, the method comprising:

[0063] Based on the students' question sets, a knowledge tracking model that supports dynamic context length is constructed;

[0064] Based on the knowledge tracking model supporting dynamic context length, the prediction result of the final student answer is obtained.

[0065] In this embodiment, if Figure 2 As shown, the knowledge tracking model supporting dynamic context length includes: student learning behavior representation module, knowledge state extraction module, directional weighted offset module, and student answer prediction module;

[0066] The student learning behavior representation module is used to describe the student's learning behavior pattern in detail by constructing a multi-dimensional feature embedding space based on representation learning theory;

[0067] The knowledge state extraction module is used to capture the deep student knowledge state based on the multi-head self-attention mechanism and dot product attention;

[0068] Directional weighted offset module, which is used to dynamically adjust the attention score matrix to achieve dynamic adaptation of different context lengths by introducing a directional weighted penalty function matrix;

[0069] The student answer prediction module is used to project the knowledge state characteristics from high dimension to low dimension with a two-layer nonlinear fully connected network as the core, optimize the binary cross entropy loss function, dynamically adjust the model parameters, and obtain the final prediction result.

[0070] In this embodiment, based on the representation learning theory, a multi-dimensional feature embedding space is constructed to finely describe the learning behavior patterns of students, including:

[0071] For student S, the historical learning behavior includes T interaction processes, each of which is expressed as t {c​t}, r t >, where t∈T, specifically, q t represents the question answered by the student at time step t, c t represents the question q answered by the student at time step t t The corresponding knowledge point. If a question corresponds to multiple knowledge points, then it is the set of multiple knowledge points {c t}, r t Indicates that the student answers question q at time step t t The answer is r t =1 means the student answered correctly, r t =0 means the student answered incorrectly;

[0072] For the questions answered by students, the encoder E(·) is used to perform feature mapping on the question sequence to obtain a high-dimensional vector representation of question-level perception:

[0073]

[0074] in, is the level of questions that can be trained, is the average feature representation of all knowledge points corresponding to the question, is the unique hot encoding of the knowledge point corresponding to the question, ⊙ represents bitwise multiplication, Represents bitwise addition;

[0075] For the student’s answer, the encoder is used to model the high-order interaction characteristics between the question knowledge points and the student’s answer through the embedding layer, thereby generating a joint question-answer interaction representation:

[0076]

[0077] Among them, v(c t , r t ) represents the average feature representation of all knowledge points corresponding to the question of the joint student answers, e(c t , r t ) represents the unique hot encoding of the knowledge points corresponding to the joint students’ corresponding questions.

[0078] The student learning behavior representation module effectively enhances the representation ability of feature embedding through the question encoder and answer encoder, providing more accurate feature input for subsequent knowledge state extraction.

[0079] In this embodiment, the deep-level student knowledge state captured based on the multi-head self-attention mechanism and dot product attention includes:

[0080] Represent the high-dimensional vector x of question-level perception tAnd the joint question-answer interaction representation yt is used as the query vector Q, key vector K and value vector V respectively, that is:

[0081] Q = x t+1 ; K = {x1, ..., x t}; V = {y1, ..., y t};

[0082] A multi-head attention mechanism integrating scaled dot-product attention is used to extract the knowledge state h from the student's historical answer sequence. t+1 :

[0083]

[0084] h t+1 =Concat(Head1, Head2, .., Head n );

[0085] Among them, Head represents the multi-head representation in the multi-head attention mechanism, Softmax represents the softmax function, represents the transpose of the key vector, represents the scaling factor of the scaled dot product attention, and d is the dimension of the key vector.

[0086] This module makes full use of the parallel computing power and hierarchical feature extraction advantages of the attention mechanism, combined with the ability of scaled dot product attention in encoding the temporal dependencies and contextual embedding characteristics inherent in student interaction data, while the multi-head attention structure further captures multi-dimensional high-order feature interaction relationships.

[0087] In this embodiment, the dynamic adaptation of different context lengths is achieved by dynamically adjusting the attention score matrix by introducing a directional weighted penalty function matrix, including:

[0088] The attention score matrix is ​​calculated using the cosine similarity between the query vector and the key vector:

[0089]

[0090] The attention score matrix is ​​dynamically adjusted through the directional penalty function matrix F:

[0091]

[0092] Among them, f ij is the element of the directional penalty function matrix F, i and j are the indices of the attention score matrix elements, that is, the distance between the question currently answered by the student and the history question. The larger the difference between i and j, the longer the student's history answer sequence. δ is the directional coefficient that can be obtained through training. The positive and negative values ​​represent the positive and negative directions of the penalty. is a long representation with directionally weighted penalties, It is a knowledge state that has been directionally weighted penalized.

[0093] By dynamically adjusting the attention score matrix through the directional penalty function matrix, the scaling dot product attention mechanism can be enhanced to capture the dynamic attenuation of students' knowledge states and realize dynamic support of context.

[0094] In this embodiment, a two-layer nonlinear fully connected network is used as the core to project the knowledge state features from high dimension to low dimension, optimize the binary cross entropy loss function, and dynamically adjust the model parameters to obtain the final prediction results including:

[0095] With a two-layer nonlinear fully connected network as the core, the knowledge state is realized The high-dimensional to low-dimensional projection of Optimize and dynamically adjust model parameters to obtain the final prediction results of students’ answers

[0096]

[0097] Among them, σ represents the Sigmoid function, k1, k2, W1, W2 are learnable parameters, and r t+1 The actual answer results for the students.

[0098] The technical solution of the present invention,

[0099] Through the question encoder and answer encoder, the student's historical answer sequence is mapped into high-dimensional feature embedding to generate question-level perception vectors and joint question-answer interaction representations.

[0100] The attention score matrix is ​​dynamically adjusted based on the directional penalty function matrix to enhance the model's ability to adapt to dynamic changes in context length.

[0101] Multi-head self-attention mechanism and scaled dot-product attention are used to extract time-sensitive and context-dependent knowledge states from student interaction data.

[0102] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0103] It should be noted that some embodiments of the present disclosure are described above. Other embodiments are within the scope of the attached claims. In some cases, it should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention. The actions or steps recorded in the claims can be performed in an order different from that in the above embodiment and still achieve the desired results. In addition, the process depicted in the accompanying drawings does not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0104] Embodiment 2

[0105] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments and methods, the present invention further discloses a knowledge tracking system supporting dynamic context, wherein the system is used to implement any of the methods, construction modules and prediction modules;

[0106] A building module, used to construct a knowledge tracking model supporting dynamic context length based on the student's question set;

[0107] The prediction module is used to obtain the prediction result of the final student answer based on the knowledge tracking model supporting dynamic context length.

[0108] The system of the above embodiment is used to implement a corresponding knowledge tracking method supporting dynamic context in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0109] It should be noted that the above-mentioned knowledge tracking system supporting dynamic context is embodied in the form of functional units. The term "module" here can be implemented in the form of software and / or hardware, and is not specifically limited to this.

[0110] For example, a "module" may be a software program, a hardware circuit, or a combination of the two that implements the above functions. The hardware circuit may include an application specific integrated circuit (ASIC), an electronic circuit, a processor (such as a shared processor, a dedicated processor, or a group processor, etc.) and a memory for executing one or more software or firmware programs, a combined logic circuit, and / or other suitable components that support the described functions.

[0111] Embodiment 3

[0112] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments, this embodiment also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a knowledge tracking method supporting dynamic context as described in any of the above embodiments is implemented.

[0113] Figure 3 A more specific schematic diagram of the hardware structure of an electronic device provided in this embodiment is shown, and the device may include: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040 are connected to each other through the bus 1050 in the device.

[0114] The processor 1010 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0115] The memory 1020 may be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 may store an operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program codes are stored in the memory 1020 and are called and executed by the processor 1010.

[0116] The input / output interface 1030 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0117] The communication interface 1040 is used to connect a communication module (not shown in the figure) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB (Universal Serial Bus), network cable, etc.), or through a wireless mode (such as mobile network, WIFI (Wireless Fidelity), Bluetooth, etc.).

[0118] The bus 1050 includes a path that transmits information between the various components of the device (eg, the processor 1010, the memory 1020, the input / output interface 1030, and the communication interface 1040).

[0119] It should be noted that, although the above device only shows the processor 1010, the memory 1020, the input / output interface 1030, the communication interface 1040 and the bus 1050, in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include the components necessary for implementing the embodiments of the present specification, and does not necessarily include all the components shown in the figure.

[0120] The system of the above embodiment is used to implement a corresponding knowledge tracking method supporting dynamic context in any of the above embodiments, and has the beneficial effects of the corresponding method embodiment, which will not be repeated here.

[0121] Embodiment 4

[0122] Based on the same inventive concept, corresponding to any of the above-mentioned embodiments, this embodiment also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute a knowledge tracking method supporting dynamic context as described in any of the above embodiments.

[0123] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.

[0124] The computer instructions stored in the storage medium of the above embodiment are used to enable the computer to execute a knowledge tracking method supporting dynamic context as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0125] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to imply that the scope of the present disclosure (including the claims) is limited to these examples. Based on the concept of the present disclosure, the technical features in the above embodiments or different embodiments may be combined, the steps may be implemented in any order, and there are many other variations of the different aspects of the embodiments of the present disclosure as described above, which are not provided in detail for the sake of simplicity.

[0126] In addition, to simplify the description and discussion, and in order not to make the embodiments of the present disclosure difficult to understand, the known power / ground connections to the integrated circuit (IC) chips and other components may or may not be shown in the provided figures. In addition, the device can be shown in the form of a block diagram to avoid making the embodiments of the present disclosure difficult to understand, and this also takes into account the fact that the details of the implementation of these block diagram devices are highly dependent on the platform on which the embodiments of the present disclosure will be implemented (that is, these details should be fully within the scope of understanding of those skilled in the art). Where specific details (e.g., circuits) are set forth to describe exemplary embodiments of the present disclosure, it is apparent to those skilled in the art that the embodiments of the present disclosure can be implemented without these specific details or with changes in these specific details. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0127] Although the present disclosure has been described in conjunction with specific embodiments of the present disclosure, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may use the embodiments discussed.

[0128] Therefore, the units of each example described in the embodiments of the present application can be implemented by 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 application.

[0129] The embodiments of the present disclosure are intended to cover all such substitutions, modifications and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the embodiments of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. A knowledge tracking method supporting dynamic context, characterized in that: The method comprises: Based on the students' question sets, a knowledge tracking model that supports dynamic context length is constructed; Based on the knowledge tracking model supporting dynamic context length, the prediction result of the final student answer is obtained.

2. The method according to claim 1, characterized in that The knowledge tracking model supporting dynamic context length includes: a student learning behavior representation module, a knowledge state extraction module, a directional weighted offset module, and a student answer prediction module; The student learning behavior representation module is used to describe the student's learning behavior pattern in detail by constructing a multi-dimensional feature embedding space based on the representation learning theory; The knowledge state extraction module is used to capture the deep-level student knowledge state based on the multi-head self-attention mechanism and dot product attention; The directional weighted offset module is used to dynamically adjust the attention score matrix by introducing a directional weighted penalty function matrix to achieve dynamic adaptation of different context lengths; The student answer prediction module is used to project the high-dimensional to low-dimensional knowledge state characteristics with a two-layer nonlinear fully connected network as the core, optimize the binary cross entropy loss function, dynamically adjust the model parameters, and obtain the final prediction result.

3. The method according to claim 2, characterized in that Based on representation learning theory, a multi-dimensional feature embedding space is constructed to finely describe students' learning behavior patterns, including: For student S, the historical learning behavior includes T interaction processes, each of which is expressed as t ,{c t }, r t >, where t∈T, specifically, q t represents the question answered by the student at time step t, c t represents the question q answered by the student at time step t t The corresponding knowledge point. If a question corresponds to multiple knowledge points, then it is the set of multiple knowledge points {c t }, r t Indicates that the student answers question q at time step t t The answer is r t =1 means the student answered correctly, r t =0 means the student answered incorrectly;​ For the questions answered by students, the encoder E(·) is used to perform feature mapping on the question sequence to obtain a high-dimensional vector representation of question-level perception: in, is the level of questions that can be trained, is the average feature representation of all knowledge points corresponding to the question, is the unique hot encoding of the knowledge point corresponding to the question, ⊙ represents bitwise multiplication, Represents bitwise addition; For the student’s answer, the encoder is used to model the high-order interaction characteristics between the question knowledge points and the student’s answer through the embedding layer, thereby generating a joint question-answer interaction representation: Among them, v(c t , r t ) represents the average feature representation of all knowledge points corresponding to the question of the joint student answers, e(c t , r t ) represents the unique hot encoding of the knowledge points corresponding to the joint students’ corresponding questions.

4. The method according to claim 2, characterized in that: Capturing the deep student knowledge state based on multi-head self-attention mechanism and dot product attention includes: Represent the high-dimensional vector x of question-level perception t And the joint question-answer interaction representation yt is used as the query vector Q, key vector K and value vector V respectively, that is: Q=x t+1 ;K={x1,...,x t };V={y1,...,y t }; A multi-head attention mechanism integrating scaled dot-product attention is used to extract the knowledge state h from the student's historical answer sequence. t+1 : h t+1 =Concat(Head1,Head2,...,Head n ); Among them, Head represents the multi-head representation in the multi-head attention mechanism, Softmax represents the softmax function, and K T represents the transpose of the key vector, represents the scaling factor of the scaled dot product attention, and d is the dimension of the key vector.

5. The method according to claim 4, characterized in that By introducing a directional weighted penalty function matrix to dynamically adjust the attention score matrix to achieve dynamic adaptation of different context lengths, including: The attention score matrix is ​​calculated using the cosine similarity between the query vector and the key vector: The attention score matrix is ​​dynamically adjusted through the directional penalty function matrix F: Among them, f ij is the element of the directional penalty function matrix F, i and j are the indices of the attention score matrix elements, δ is the directional coefficient that can be obtained through training, and the positive and negative values ​​represent the positive and negative directions of the penalty. is a long representation with a directionally weighted penalty, It is a knowledge state that has been directionally weighted penalized.

6. The method according to claim 5, characterized in that With a two-layer nonlinear fully connected network as the core, the high-dimensional to low-dimensional projection of knowledge state features is performed, the binary cross entropy loss function is optimized, and the model parameters are dynamically adjusted to obtain the final prediction results including: With a two-layer nonlinear fully connected network as the core, the knowledge state is realized The high-dimensional to low-dimensional projection of Optimize and dynamically adjust model parameters to obtain the final prediction results of students’ answers Among them, σ represents the Sigmoid function, k1, k2, W1, W2 are learnable parameters, and r t+1 The actual answer results for the students.

7. A knowledge tracking system supporting dynamic context, the system being used to implement the method of any one of claims 1 to 6, characterized in that: The system comprises: a construction module and a prediction module; The building module is used to build a knowledge tracking model supporting dynamic context length based on the student's question set; The prediction module is used to obtain the prediction result of the final student answer based on the knowledge tracking model supporting dynamic context length.

8. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed, the method according to any one of claims 1 to 6 is implemented.

Images