Emotion enhancement knowledge tracking method and system based on graph neural network, and storage medium

Abstract

The invention discloses an emotion enhancement knowledge tracking method and system based on a graph neural network and a storage medium, and the method comprises the steps: obtaining question embedding, skill embedding and answer embedding according to the exercise interaction information of a learner, constructing a question-skill heterogeneous graph, and obtaining emotion embedding according to the emotion state information of the learner; performing feature extraction on the problem-skill heterogeneous graph through a trained graph convolutional network model to obtain problem-skill embedding; obtaining a final exercise embedding according to the question-skill embedding, the answer embedding and the emotion embedding; inputting the final exercise embedding and emotion embedding of each time step into the trained knowledge state evolution model, and outputting the current knowledge state of the learner; and predicting an answer result of the to-be-predicted question according to the current knowledge state of the learner. The knowledge tracking model is constructed in an emotion embedding enhancement mode, the prediction accuracy is improved, and the method can be widely applied to the technical field of intelligent education.

Description

Technical Field

[0001] The present invention relates to the field of intelligent education technology, and in particular to a method, system and storage medium for tracking emotion-enhanced knowledge based on a graph neural network. Background Art

[0002] Knowledge tracing is a method for modeling changes in students' knowledge status and is the basis for personalized learning. Most of the current mainstream emotion-enhanced knowledge tracing methods based on graph neural networks do not consider the impact of learners' emotional factors on the results when they answer. However, emotional factors can greatly affect learners' learning performance. Ignoring the impact of emotional factors will cause the knowledge tracing task to model the learners' non-real level, reducing the effectiveness of the model. Summary of the invention

[0003] The purpose of the present invention is to solve one of the technical problems existing in the prior art to at least a certain extent.

[0004] To this end, one purpose of an embodiment of the present invention is to provide a sentiment-enhanced knowledge tracking method based on graph neural network, which jointly models the learner's knowledge state and emotional state by combining the graph convolution network model (GCN) and the Transformer model, comprehensively considers the learner's emotional state information and the time series information of knowledge evolution, and effectively improves the prediction accuracy of the knowledge tracking model.

[0005] Another object of an embodiment of the present invention is to provide a sentiment-enhanced knowledge tracking method system based on graph neural network.

[0006] In order to achieve the above technical objectives, the technical solutions adopted by the embodiments of the present invention include:

[0007] In a first aspect, an embodiment of the present invention provides a method for tracking sentiment-enhanced knowledge based on a graph neural network, characterized in that the method comprises the following steps:

[0008] Obtain learners' exercise interaction information and emotional state information;

[0009] Obtaining question embedding, skill embedding and answer embedding according to the exercise interaction information, and obtaining emotion embedding according to the emotion state information;

[0010] Constructing a problem-skill heterogeneous graph according to the problem embedding and the skill embedding, and extracting features of the problem-skill heterogeneous graph through a trained graph convolutional network model to obtain a problem-skill embedding;

[0011] Obtaining a final exercise embedding according to the question-skill embedding, the answer embedding, and the emotion embedding;

[0012] Inputting the final exercise embedding and the sentiment embedding at each time step into the trained knowledge state evolution model, and outputting the current knowledge state of the learner;

[0013] An answer result of the question to be predicted is generated according to the current knowledge state of the learner and the question to be predicted.

[0014] Furthermore, the exercise interaction information includes the learner's student set, question set, skill set, answer set and answer timestamp information.

[0015] Further, obtaining question embedding, skill embedding, and answer embedding according to the exercise interaction information, and obtaining emotion embedding according to the emotion state information includes:

[0016] Embedding the question set, the skill set, and the answer set to obtain the question embedding, the skill embedding, and the answer embedding;

[0017] The emotional state information is discretized and standardized to obtain the emotional embedding.

[0018] Further, constructing a problem-skill heterogeneous graph according to the problem embedding and the skill embedding includes:

[0019] Constructing a problem-skill bipartite graph according to the problem set and the skill set;

[0020] According to the problem-skill bipartite graph, the relationship between the problem set and the skill set is analyzed by an association rule mining algorithm to construct the problem-skill heterogeneous graph.

[0021] Further, obtaining a final exercise embedding according to the question-skill embedding, the answer embedding and the emotion embedding includes:

[0022] Aggregating the question-skill embedding, the answer embedding, and the sentiment embedding to obtain an aggregated embedding;

[0023] A nonlinear transformation is performed on the aggregated embedding to obtain the final training embedding.

[0024] Further, generating an answer result of the question to be predicted according to the current knowledge state of the learner and the question to be predicted includes:

[0025] Processing the problem to be predicted by using the trained graph convolutional network model to obtain an embedding of the problem to be predicted;

[0026] The current knowledge state and the problem to be predicted are embedded in a fully connected network and an activation function for combination and nonlinear transformation to obtain an answer to the problem to be predicted.

[0027] Furthermore, the graph convolutional network model and the knowledge state evolution model are trained by the following steps:

[0028] Compare the predicted answer result of the question to be predicted with the actual answer result by using a loss function to obtain a loss value;

[0029] According to the loss value, the parameters of the graph convolutional network model and the knowledge state evolution model are adjusted through a back propagation algorithm.

[0030] In a second aspect, an embodiment of the present invention provides a sentiment-enhanced knowledge tracking system based on a graph neural network, characterized in that it includes:

[0031] Data collection module, used to obtain learners' exercise interaction information and emotional state information;

[0032] A data preprocessing module, used to obtain question embedding, skill embedding and answer embedding according to the exercise interaction information, and obtain emotion embedding according to the emotion state information;

[0033] A feature extraction module, used to construct a problem-skill heterogeneous graph according to the problem embedding and the skill embedding, and extract features from the problem-skill heterogeneous graph through a trained graph convolutional network model to obtain a problem-skill embedding;

[0034] A feature fusion module, used to obtain a final exercise embedding according to the question-skill embedding, the answer embedding and the sentiment embedding;

[0035] A knowledge state prediction module, used for inputting the final exercise embedding and the sentiment embedding of each time step into the trained knowledge state evolution model, and outputting the current knowledge state of the learner;

[0036] The answer result prediction module is used to generate an answer result of the question to be predicted based on the current knowledge state of the learner and the question to be predicted.

[0037] In a third aspect, an embodiment of the present invention provides a device, including:

[0038] at least one processor;

[0039] at least one memory for storing at least one program;

[0040] When the at least one program is executed by the at least one processor, the at least one processor implements the above-mentioned sentiment-enhanced knowledge tracking method based on graph neural network.

[0041] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, which stores a program executable by a processor, and the program executable by the processor is used to execute the above-mentioned sentiment-enhanced knowledge tracking method based on graph neural network when executed by the processor.

[0042] The advantages and beneficial effects of the present invention will be partly given in the following description, partly become apparent from the following description, or be understood through the practice of the present invention:

[0043] The present invention constructs a problem-skill heterogeneous graph by analyzing the relationship between the problem set and the skill set, uses a graph convolutional neural network to extract high-order features of the problem-skill heterogeneous graph, and obtains the learner's knowledge state characteristics that comprehensively consider the emotional state information after fusing the answer embedding and the emotion embedding. The knowledge state characteristics and the emotional state characteristics of each time step are used as the input of the Transformer model, completing the modeling of the knowledge evolution model based on the self-attention mechanism, which can perform a comprehensive dynamic evaluation of the learner's knowledge state and emotional fluctuations, can effectively predict the learner's answer situation, and improve the prediction accuracy of the knowledge tracking model. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A step diagram of a method for tracking sentiment-enhanced knowledge based on a graph neural network provided by an embodiment of the present invention;

[0045] Figure 2 A schematic diagram of a technical solution flow of a method for tracking emotion-enhanced knowledge based on a graph neural network provided by an embodiment of the present invention;

[0046] Figure 3 A schematic diagram of a sentiment-enhanced knowledge tracking system based on a graph neural network provided by an embodiment of the present invention;

[0047] Figure 4 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0049] In the description of the present invention, the meaning of "a plurality" is two or more than two. If there is a description of "a first" or "a second", it is only used for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features. In addition, unless otherwise defined, all technical and scientific terms used in this document have the same meaning as those commonly understood by those skilled in the art.

[0050] Figure 1 A step diagram of a method for tracking sentiment-enhanced knowledge based on a graph neural network provided by an embodiment of the present invention, referring to Figure 1 The embodiment of the present invention provides a method for tracking emotion-enhanced knowledge based on a graph neural network, which is characterized by comprising the following steps:

[0051] Obtain learners' exercise interaction information and emotional state information;

[0052] Based on the exercise interaction information, we get question embedding, skill embedding and answer embedding, and based on the emotional state information, we get emotional embedding;

[0053] Construct a problem-skill heterogeneous graph based on problem embedding and skill embedding, extract features from the problem-skill heterogeneous graph through the trained graph convolutional network model, and obtain problem-skill embedding;

[0054] The final practice embedding is obtained based on the question-skill embedding, answer embedding, and emotion embedding;

[0055] Input the final exercise embedding and sentiment embedding of each time step into the trained knowledge state evolution model and output the learner’s current knowledge state;

[0056] Generate answers to questions to be predicted based on the learner's current knowledge status and the questions to be predicted.

[0057] Specifically, specifically, Figure 2 A schematic diagram of a technical solution flow chart of a method for tracking emotion-enhanced knowledge based on a graph neural network provided in this embodiment. Figure 2The implementation process of the technical solution of this embodiment is introduced. In this implementation, by analyzing the students' practice data, information such as the student set, question set, skill set, answer set, and answer timestamp information can be obtained, and the students' emotional state information can be obtained by asking questions in the questionnaire and combining computer vision before answering the questions. The timestamp information of the emotional state and the timestamp information of the answer are aligned to obtain the emotional information corresponding to each answer data.

[0058] In this embodiment, after obtaining the learner's exercise interaction information and emotional state information, the embedding layer processes to obtain the embedded representation of the student's questions, skills, answers and emotions in the set (q i ,c i ,a i ,e i ). There are four embedding matrices in total:

[0059]

[0060] Among them, |Q| represents the size of the problem set, that is, the total number of all problems, |S| represents the size of the skill set, that is, the total number of all skills, |E| represents the number of types of emotional states, and the second dimension d represents the feature dimension of embedding. The data generated by the learner in the learning process is embedded and the corresponding embedding representation (q i ,c i ,a i ,e i ).

[0061] After embedding processing, we get the student set Question Collection Skill Set and the set of emotional states Use N s , N q , N c and E s represents the number of students, questions, skills, and emotions. Assume that at time t, the student’s knowledge state is h t ,h t Indicates the degree to which students have mastered each piece of knowledge.

[0062] In this embodiment, after the embedded representation of each indicator is obtained, the knowledge association matrix QC is constructed according to the known prior relationship between the problem and the skill. QC is a matrix containing only 0 and 1. i Contains knowledge concept c k When QC jk =1, otherwise QC jk= 0, j and k represent the number of rows and columns corresponding to QC respectively. By aligning the timestamp information of the emotional state with the timestamp information of the answer, the emotional state of the student corresponding to each answer data can be obtained. i Therefore, the cognitive data of students in the learning process can be expressed as a set of historical cognitive records:

[0063] {(q1,c1,a1),e1,(q2,c2,a2),e2,…,(q i ,c i ,a i ),e i ,…,(q t ,c t ,,a t ),e t}

[0064] Among them, a t ∈{0,1} indicates whether the student answered question q correctly at time t t .

[0065] In this embodiment, from the perspective of reasoning, whether a student can correctly answer a new question depends on his or her mastery of relevant skills and the characteristics of the question. If a student has solved a similar problem before, then he or she is more likely to answer a new question correctly. Since the question-skill relationship graph is bidirectional, the first-hop neighbor of a problem should be its corresponding skill, and the second-hop neighbor should be other problems that share the same skill. Therefore, the knowledge association matrix QC can only capture the direct relationship between problems and skills, but cannot capture the deeper implicit information between problems and skills. To this end, in this embodiment, based on the constructed knowledge association matrix QC, the relationship between problems and skills (knowledge concepts) is modeled as a heterogeneous graph, and the node features are aggregated layer by layer through the graph convolutional network (GCN) to obtain a richer knowledge representation. The graph convolutional network is used to update the problem embeddings and skill embeddings of related skills and problems through embedding propagation, thereby extracting high-order feature information.

[0066] The graph convolutional network stacks multiple graph convolutional layers to encode high-order neighboring information. In each layer, the node representation can be updated by embedding itself and its neighboring nodes. The representation of node i in the graph is denoted by x i (x i It can represent the skill embedding c i Or question embedding q i , its neighbor node set is recorded as N i ), then the formula of the l-th layer GCN is:

[0067]

[0068] in, is the representation of node i at layer l, N i is the neighbor set of node i, W (l) and b (l) are the weight matrix and bias of the lth layer respectively, σ represents the activation function, which can be the ReLU function in this embodiment.

[0069] In this embodiment, after extracting high-order knowledge state features through the above processing, it is necessary to subsequently integrate the knowledge state features with the emotion embedding vector e t 、Answer embedded in a t And the question embedding obtained by graph convolutional network Fusion is performed to enhance the model's ability to perceive changes in students' emotions. Specifically, the question, answer, and emotion are first fused through the feature fusion module to obtain the aggregate embedding of the three. Then, the aggregate embedding information is passed through a layer of nonlinear neural network to obtain the final exercise embedding information. The calculation formula for the whole process is as follows:

[0070]

[0071] Among them, x t is the output of the fusion module, which represents the comprehensive vector of students considering emotional factors. is the question embedding obtained through the graph convolutional network (GCN), a t is the answer embedding, e t is the sentiment embedding, W1 and b1 are the weight matrix and bias vector used to integrate and map different embeddings to the same dimension.

[0072] In this embodiment, after obtaining the final exercise embedding information, a knowledge state evolution model based on the self-attention mechanism is constructed based on the Transformer model.

[0073] In order to overcome the limitations of traditional LSTM in long sequence modeling, by choosing to use the Transformer model to evolve the students' knowledge state, we can overcome the limitations of traditional LSTM in long sequence modeling and model the long-term dependencies of students' learning sequences.

[0074] In the self-attention module, the input interaction sequence X is first transformed linearly to obtain the query matrix Q, key matrix K and value matrix V, and then the attention weight A is calculated:

[0075]

[0076] Among them, d k Represents the dimension of the key vector, which is used to scale the attention score. Through parallel computation of multiple attention heads, Transformer can effectively capture long-term dependencies and complex relationships.

[0077] In combination with the emotional state information, by adding the learner’s emotional state embedding at each time step to the input of the Transformer, an emotionally enhanced knowledge state representation can be formed:

[0078] h t = Transformer([x,x2,…,x t ,e t ])

[0079] Among them, e t It represents the emotional state embedding of the learner at time step t. Through the fusion of emotional states, the understanding of the learner’s current knowledge state is enhanced.

[0080] In this embodiment, after obtaining the learner's current knowledge state h t Then, using the current state and the problem to be predicted Specifically, a fully connected network is used and activated by a sigmoid function to predict the correct probability of the learner's answer to a new question:

[0081]

[0082] Where W out is the weight matrix of the output layer, b out is the bias term.

[0083] In this embodiment, after completing the joint modeling based on the graph convolutional network model and the Transformer model, the cross entropy loss function is used to train the model by utilizing the learner's historical learning data and emotional state:

[0084]

[0085] where y i is the actual label, is the prediction probability of the model. By optimizing the loss function, the parameters in the graph neural network and Transformer module are updated to improve the prediction accuracy and generalization ability.

[0086] It can be recognized that this embodiment constructs a problem-skill heterogeneous graph by analyzing the relationship between the problem set and the skill set, uses a graph convolutional neural network to extract high-order features of the problem-skill heterogeneous graph, and integrates the answer embedding and emotion embedding to obtain the learner's knowledge state characteristics that comprehensively consider the emotional state information. The knowledge state characteristics and emotional state characteristics of each time step are used as inputs to the Transformer model, completing the modeling of the knowledge evolution model based on the self-attention mechanism, which can perform a comprehensive dynamic evaluation of the learner's knowledge state and emotional fluctuations, can effectively predict the learner's answer situation, and improve the prediction accuracy of the knowledge tracking model.

[0087] In some optional embodiments, the exercise interaction information includes the learner's student set, question set, skill set, answer set, and answer timestamp information.

[0088] Specifically, in this embodiment, the student set is mainly used to distinguish different learners and ensure that the model can track knowledge based on the historical learning records of different learners; the question set and skill set include all questions and all corresponding knowledge points, and the correspondence between the two can be obtained through pre-annotation for the subsequent construction of question-skill heterogeneous graphs; the answer set is used to obtain the learners' answers, and the deviation between the predicted answer results and the actual answer results is used to calculate the loss value, and the parameters of the model are updated through the back propagation algorithm; the answer timestamp information is used to align the students' emotional state with the answer results, providing data support for subsequent time series modeling.

[0089] In some optional embodiments, obtaining question embedding, skill embedding, and answer embedding according to exercise interaction information, and obtaining emotion embedding according to emotion state information includes:

[0090] Embedding the question set, skill set and answer set to obtain question embedding, skill embedding and answer embedding;

[0091] The emotional state information is discretized and standardized to obtain emotional embedding.

[0092] Specifically, in this embodiment, the embedding processing method is mainly to map the discrete high-dimensional data in each set to a low-dimensional vector space, such as mapping discrete question numbers, skill numbers, and answer results into corresponding vectors through an embedding matrix. As for the emotional state information, each emotional state type can be divided into several possible types in advance, such as anxiety, calmness, etc. The basis for the division is that there are obvious substantive differences between the categories, which are easy to distinguish, and different types of emotional states may have a more obvious impact on the answering situation. After the categories are divided, the emotional state is mapped to discrete numerical data and then standardized to ensure the uniformity of the data range.

[0093] In some optional embodiments, constructing a problem-skill heterogeneous graph according to the problem embedding and the skill embedding includes:

[0094] Construct a problem-skill bipartite graph based on the problem set and skill set;

[0095] According to the problem-skill bipartite graph, the relationship between the problem set and the skill set is analyzed through the association rule mining algorithm to construct a problem-skill heterogeneous graph.

[0096] In some optional embodiments, obtaining the final exercise embedding according to the question-skill embedding, the answer embedding, and the emotion embedding includes:

[0097] Aggregate question-skill embedding, answer embedding, and sentiment embedding to get aggregate embedding;

[0098] Perform a nonlinear transformation on the aggregated embedding to obtain the final training embedding.

[0099] In some optional embodiments, generating an answer result of the question to be predicted according to the current knowledge state of the learner and the question to be predicted includes:

[0100] The problem to be predicted is processed through the trained graph convolutional network model to obtain the embedding of the problem to be predicted;

[0101] The current knowledge state and the question to be predicted are embedded in a fully connected network and an activation function for combination and nonlinear transformation to obtain the answer to the question to be predicted.

[0102] In some optional embodiments, the graph convolutional network model and the knowledge state evolution model are trained by the following steps:

[0103] The loss function is used to compare the predicted answer result with the actual answer result to obtain the loss value.

[0104] According to the loss value, the parameters of the graph convolutional network model and the knowledge state evolution model are adjusted through the back-propagation algorithm.

[0105] Reference Figure 3 The embodiment of the present invention provides a sentiment enhancement knowledge tracking system based on graph neural network, which is characterized by comprising:

[0106] Data collection module, used to obtain learners' exercise interaction information and emotional state information;

[0107] The data preprocessing module is used to obtain question embedding, skill embedding and answer embedding based on the exercise interaction information, and to obtain emotion embedding based on the emotion state information;

[0108] The feature extraction module is used to construct a problem-skill heterogeneous graph based on problem embedding and skill embedding. The trained graph convolutional network model is used to extract features from the problem-skill heterogeneous graph to obtain the problem-skill embedding.

[0109] Feature fusion module, used to obtain the final exercise embedding based on question-skill embedding, answer embedding and sentiment embedding;

[0110] The knowledge state prediction module is used to input the final exercise embedding and sentiment embedding of each time step into the trained knowledge state evolution model and output the current knowledge state of the learner;

[0111] The answer result prediction module is used to generate the answer result of the question to be predicted based on the learner's current knowledge status and the question to be predicted.

[0112] Reference Figure 4 , an embodiment of the present invention provides a computer device, including:

[0113] at least one processor;

[0114] at least one memory for storing at least one program;

[0115] When the at least one program is executed by the at least one processor, the at least one processor implements the emotion-enhanced knowledge tracking method based on graph neural network.

[0116] The contents of the above method embodiments are all applicable to the present device embodiments. The functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0117] An embodiment of the present invention also provides a computer-readable storage medium, which stores a program executable by a processor. When the program executable by the processor is executed by the processor, it is used to execute the above-mentioned emotion-enhanced knowledge tracking method based on graph neural network.

[0118] A computer-readable storage medium of an embodiment of the present invention can execute a sentiment-enhanced knowledge tracking method based on a graph neural network provided by an embodiment of the method of the present invention, can execute any combination of implementation steps of the method embodiment, and has the corresponding functions and beneficial effects of the method.

[0119] The embodiment of the present invention also discloses a computer program product or a computer program, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 1The sentiment-enhanced knowledge tracking method based on graph neural network is shown.

[0120] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the above-mentioned boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented herein. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part for which is described as a larger operation is performed independently.

[0121] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the above-mentioned functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of the various functional modules in the device disclosed herein, the actual implementation of the module will be understood within the conventional skills of the engineer. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0122] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the above methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0123] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0124] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and editable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the above-mentioned program is printed, since the above-mentioned program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or processing in other suitable ways as necessary, and then stored in a computer memory.

[0125] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0126] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0127] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0128] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the above embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A sentiment-enhanced knowledge tracking method based on graph neural network, characterized in that: The following steps are involved: Obtain learners' exercise interaction information and emotional state information; Obtaining question embedding, skill embedding and answer embedding according to the exercise interaction information, and obtaining emotion embedding according to the emotion state information; Constructing a problem-skill heterogeneous graph according to the problem embedding and the skill embedding, and extracting features of the problem-skill heterogeneous graph through a trained graph convolutional network model to obtain a problem-skill embedding; Obtaining a final exercise embedding according to the question-skill embedding, the answer embedding, and the emotion embedding; Inputting the final exercise embedding and the sentiment embedding at each time step into the trained knowledge state evolution model, and outputting the current knowledge state of the learner; An answer result of the question to be predicted is generated according to the current knowledge state of the learner and the question to be predicted.

2. According to the method of claim 1, the emotion-enhanced knowledge tracking method based on graph neural network is characterized in that: The exercise interaction information includes the learner's student set, question set, skill set, answer set, and answer timestamp information.

3. The emotion-enhanced knowledge tracking method based on graph neural network according to claim 2 is characterized in that: The step of obtaining question embedding, skill embedding, and answer embedding according to the exercise interaction information, and obtaining emotion embedding according to the emotion state information comprises: Embedding the question set, the skill set, and the answer set to obtain the question embedding, the skill embedding, and the answer embedding; The emotional state information is discretized and standardized to obtain the emotional embedding.

4. The emotion-enhanced knowledge tracking method based on graph neural network according to claim 2 is characterized in that: The constructing a problem-skill heterogeneous graph according to the problem embedding and the skill embedding comprises: Constructing a problem-skill bipartite graph according to the problem set and the skill set; According to the problem-skill bipartite graph, the relationship between the problem set and the skill set is analyzed by an association rule mining algorithm to construct the problem-skill heterogeneous graph.

5. The emotion-enhanced knowledge tracking method based on graph neural network according to claim 1 is characterized in that: The obtaining of the final exercise embedding according to the question-skill embedding, the answer embedding and the emotion embedding comprises: Aggregating the question-skill embedding, the answer embedding, and the sentiment embedding to obtain an aggregated embedding; A nonlinear transformation is performed on the aggregated embedding to obtain the final training embedding.

6. The emotion-enhanced knowledge tracking method based on graph neural network according to claim 1 is characterized in that: The generating of the answer result of the question to be predicted according to the current knowledge state of the learner and the question to be predicted includes: Processing the problem to be predicted by using the trained graph convolutional network model to obtain an embedding of the problem to be predicted; The current knowledge state and the problem to be predicted are embedded in a fully connected network and an activation function for combination and nonlinear transformation to obtain an answer to the problem to be predicted.

7. The emotion-enhanced knowledge tracking method based on graph neural network according to claim 1 is characterized in that: The graph convolutional network model and the knowledge state evolution model are trained by the following steps: Compare the predicted answer result of the question to be predicted with the actual answer result by using a loss function to obtain a loss value; According to the loss value, the parameters of the graph convolutional network model and the knowledge state evolution model are adjusted through a back propagation algorithm.

8. A sentiment-enhanced knowledge tracking system based on graph neural network, characterized in that: include: Data collection module, used to obtain learners' exercise interaction information and emotional state information; A data preprocessing module, used to obtain question embedding, skill embedding and answer embedding according to the exercise interaction information, and obtain emotion embedding according to the emotion state information; A feature extraction module, used to construct a problem-skill heterogeneous graph according to the problem embedding and the skill embedding, and extract features from the problem-skill heterogeneous graph through a trained graph convolutional network model to obtain a problem-skill embedding; A feature fusion module, used to obtain a final exercise embedding according to the question-skill embedding, the answer embedding and the sentiment embedding; A knowledge state prediction module, used for inputting the final exercise embedding and the sentiment embedding of each time step into the trained knowledge state evolution model, and outputting the current knowledge state of the learner; The answer result prediction module is used to generate an answer result of the question to be predicted based on the current knowledge state of the learner and the question to be predicted.

9. A device, characterized in that: include: at least one processor; at least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a sentiment-enhanced knowledge tracking method based on a graph neural network as described in any one of claims 1-7.

10. A computer-readable storage medium storing a program executable by a processor, characterized in that: The processor-executable program is used to execute a sentiment-enhanced knowledge tracking method based on a graph neural network as described in any one of claims 1-7 when executed by the processor.

Images