A method, device and equipment for tracking related knowledge

Abstract

The embodiments of this specification disclose a method, apparatus and equipment for tracking associated knowledge, including: obtaining each question-answering interaction data of a user within a specified time range, the question-answering interaction data including the knowledge points answered at the corresponding time, correct or incorrect answer information, and an associated knowledge point set consisting of associated knowledge points of the answered knowledge points; inputting each question-answering interaction data into an associated knowledge tracking model for processing according to the corresponding time sequence, wherein the associated knowledge tracking model is pre-trained according to a first loss and a second loss, the first loss being determined based on prediction information of the mastery level of the specified knowledge point based on training samples and its training labels, and the second loss being determined based on prediction information of the mastery level of the associated knowledge point set of the specified knowledge point based on training samples and its training labels; through processing by the associated knowledge tracking model, the user's mastery level of the knowledge points in the predetermined knowledge point set after the specified time range is predicted.

Description

Technical Field

[0001] This specification relates to the field of machine learning technology, and in particular to a method, device, and equipment for tracking associated knowledge. Background Art

[0002] With the rapid development of AI technology, knowledge tracking has begun to play an increasingly important role in personalized user assessments in scenarios such as education and skills training.

[0003] In the existing technology, most knowledge tracking methods often ignore the correlation between knowledge points when predicting the knowledge mastery level of participating users, and are unable to effectively track the knowledge points in the questions, making it difficult to accurately predict the learning status of participating users on the overall knowledge points.

[0004] Based on this, an effective tracking method for related knowledge points is needed to better predict the learning status of participating users on the overall knowledge points. Summary of the Invention

[0005] One or more embodiments of this specification provide a method, apparatus, and device for tracking related knowledge to solve the following technical problems:

[0006] An effective tracking method for related knowledge points is now needed to better predict the learning status of participating users on the overall knowledge points.

[0007] One or more embodiments of this specification adopt the following technical solutions:

[0008] One or more embodiments of this specification provide a method for tracking associated knowledge, including:

[0009] Obtaining interactive data on each question answering by the user within a specified time range, the interactive data including the knowledge points answered at the corresponding time, correct or incorrect information about the answers, and a set of associated knowledge points consisting of associated knowledge points of the knowledge points answered;

[0010] Inputting the various question-answering interaction data into an associated knowledge tracking model for processing in a corresponding time sequence, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on prediction information of the mastery level of a specified knowledge point based on a training sample and its training label, and the second loss is determined based on prediction information of the mastery level of a set of associated knowledge points of the specified knowledge point based on the training sample and its training label;

[0011] Through the processing of the associated knowledge tracking model, the user's mastery of the knowledge points in the predetermined knowledge point set after the specified time range is predicted.

[0012] One or more embodiments of this specification provide a device for tracking associated knowledge, including:

[0013] an acquisition unit, which acquires each question-answering interaction data of the user within a specified time range, wherein the question-answering interaction data includes the knowledge point answered at the corresponding time, the correct or incorrect answer information, and a set of associated knowledge points consisting of associated knowledge points of the knowledge point answered;

[0014] an input unit that inputs the question-answering interaction data into an associated knowledge tracking model for processing in a corresponding time sequence, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on prediction information of a mastery level of a specified knowledge point based on a training sample and its training label, and the second loss is determined based on prediction information of a mastery level of a set of associated knowledge points of the specified knowledge point based on the training sample and its training label;

[0015] The prediction unit predicts the user's mastery of the knowledge points in the predetermined knowledge point set after the specified time range through the processing of the associated knowledge tracking model.

[0016] One or more embodiments of this specification provide a related knowledge tracking device, including:

[0017] at least one processor; and,

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0020] Obtaining interactive data on each question answering by the user within a specified time range, the interactive data including the knowledge points answered at the corresponding time, correct or incorrect information about the answers, and a set of associated knowledge points consisting of associated knowledge points of the knowledge points answered;

[0021] Inputting the various question-answering interaction data into an associated knowledge tracking model for processing in a corresponding time sequence, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on prediction information of the mastery level of a specified knowledge point based on a training sample and its training label, and the second loss is determined based on prediction information of the mastery level of a set of associated knowledge points of the specified knowledge point based on the training sample and its training label;

[0022] Through the processing of the associated knowledge tracking model, the user's mastery of the knowledge points in the predetermined knowledge point set after the specified time range is predicted.

[0023] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured to:

[0024] Obtaining interactive data on each question answering by the user within a specified time range, the interactive data including the knowledge points answered at the corresponding time, correct or incorrect information about the answers, and a set of associated knowledge points consisting of associated knowledge points of the knowledge points answered;

[0025] Inputting the various question-answering interaction data into an associated knowledge tracking model for processing in a corresponding time sequence, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on prediction information of the mastery level of a specified knowledge point based on a training sample and its training label, and the second loss is determined based on prediction information of the mastery level of a set of associated knowledge points of the specified knowledge point based on the training sample and its training label;

[0026] Through the processing of the associated knowledge tracking model, the user's mastery of the knowledge points in the predetermined knowledge point set after the specified time range is predicted.

[0027] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0028] 1. The embodiments of this specification can obtain the interactive data of each question answering by the user within a specified time range. The interactive data includes the knowledge point answered, the associated knowledge point set consisting of the associated knowledge points of the answered knowledge point, and the correct or incorrect answer information. When the associated knowledge tracking model processes the interactive data of each question answering, it can predict the mastery of the knowledge point based on the knowledge point dimension, and better predict the learning status of the participating users on the overall knowledge points.

[0029] 2. In the embodiment of this specification, the interactive data of each question is input into the associated knowledge tracking model for processing according to the corresponding time sequence, which can speed up the processing speed of the associated knowledge tracking model and enable the model to obtain prediction results more quickly;

[0030] 3. The associated knowledge tracking model of the embodiment of this specification is trained based on the first loss and the second loss, and the first loss is determined based on the mastery level prediction information of the specified knowledge point and its training label, and the second loss is determined based on the mastery level prediction information of the set of associated knowledge points of the specified knowledge point and its training label. This can make the trained associated knowledge tracking model correlated with the mastery level prediction information of the specified knowledge point and the mastery level prediction information of its set of associated knowledge points, thereby making the trained associated knowledge tracking model more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some of the embodiments described in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0032] Figure 1 A flowchart of a method for tracking associated knowledge provided in one or more embodiments of this specification;

[0033] Figure 2 A schematic diagram of the knowledge points and their associated knowledge point sets provided for one or more embodiments of this specification;

[0034] Figure 3 A flowchart for generating a dictionary of topic knowledge points provided in one or more embodiments of this specification;

[0035] Figure 4 A schematic diagram of the overall structure of the associated knowledge tracking model provided for one or more embodiments of this specification;

[0036] Figure 5 A schematic diagram of a problem-level online tutoring system provided for one or more embodiments of this specification;

[0037] Figure 6 A schematic diagram of a knowledge point-level online tutoring system provided in one or more embodiments of this specification;

[0038] Figure 7 A schematic diagram illustrating an example of the prediction process of the associated knowledge tracking model provided in one or more embodiments of this specification on an open source dataset;

[0039] Figure 8 A schematic diagram of the structure of an associated knowledge tracking device provided in one or more embodiments of this specification;

[0040] Figure 9A schematic diagram of the structure of an associated knowledge tracking device provided in one or more embodiments of this specification. DETAILED DESCRIPTION

[0041] The embodiments of this specification provide a method, apparatus, and device for tracking associated knowledge.

[0042] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this specification without creative work should fall within the scope of protection of this specification.

[0043] Figure 1 A flowchart of an associated knowledge tracking method provided for one or more embodiments of this specification is provided. The process can be executed by an associated knowledge tracking platform, which can be applied to online education systems to test users' mastery of knowledge points. Certain input parameters or intermediate results in the process allow manual intervention and adjustment to help improve accuracy.

[0044] The method steps of the embodiment of this specification are as follows:

[0045] S102, obtaining the user's interactive data on answering questions within a specified time range, the interactive data including the knowledge points answered at the corresponding time, correct or incorrect information of the answers, and a set of associated knowledge points consisting of associated knowledge points of the knowledge points answered.

[0046] In the embodiments of this specification, the associated knowledge points of the answered knowledge point include: one or more knowledge points in the answered question other than the answered knowledge point. For example, the interactive data of a question answering question includes the answered knowledge point of the question, the correct or incorrect answer information (whether the question is answered correctly or incorrectly), and the associated knowledge points of the answered knowledge point (including one or more associated knowledge points).

[0047] The interactive data of each answer within a specified time range can be considered as multiple answer data within a period of time. In this case, the multiple answer data are multiple answer data about the questions formed by the answered knowledge points and the set of related knowledge points.

[0048] The knowledge points answered and their associated knowledge points can be considered as different settings for different questions. As the time of answering changes, the knowledge points answered and their associated knowledge points change dynamically. A certain associated knowledge point becomes the answered knowledge point at the time it is answered. A knowledge point answered in a question at one time may become an associated knowledge point in a question at another time.

[0049] The corresponding moment of the question-answering interaction data can be considered as the moment of answering the answered knowledge point (it should be noted that the so-called moment may not be a specific and clear time point. It is mainly used to indicate the order of an action of answering a certain knowledge point in a certain question on the timeline. It is generally considered that users answer the various knowledge points contained in the same question in sequence. However, users may not answer different questions in sequence, and users may not need to answer all questions. Users may answer the same question repeatedly).

[0050] There is a corresponding order for these moments. The order of the moments mentioned here can be set according to the knowledge points answered, so that the user's learning of the overall knowledge points can be better predicted. Assuming that the knowledge points contained in a certain question are 1, 2, and 3 in sequence, then the knowledge point answered by the user at time t is knowledge point 1, and knowledge points 2 and 3 are set as related knowledge points; at time t+1 (that is, the next moment after time t, the next time the knowledge point is answered), the knowledge point answered is knowledge point 2, and knowledge points 1 and knowledge points 3 are set as related knowledge points; at time t+2, the knowledge point answered is knowledge point 3, and knowledge points 1 and knowledge points 2 are set as related knowledge points.

[0051] Furthermore, before obtaining the user's interactive data on answering questions within a specified time range, the embodiments of this specification read the original data set of questions and knowledge points from the relational database in advance. The questions in the original data set include all questions and knowledge points. The interactive data on answering questions within the specified time range may be some questions selected for each user. At this time, some of the questions may be questions corresponding to knowledge points that the user has not mastered before.

[0052] Then, the original data set is traversed to obtain the serial number of each question, and for the serial number of each question, the original data set is traversed to obtain a set of knowledge points with the serial number consistent with the serial number of each question, and then according to the correspondence between the serial number of the question and the set of knowledge points, a dictionary of question knowledge point sets is constructed, so that the answered knowledge points of the answered question and their associated knowledge point sets can be obtained by querying the dictionary of question knowledge point sets.

[0053] The above method can quickly and easily obtain the user's answered questions, the knowledge points answered for the answered questions, and the set of related knowledge points from the original data set. At the same time, the answered questions can be targeted at each user, which can more effectively help each user track related knowledge.

[0054] S104: Input each question-answering interaction data into the associated knowledge tracking model for processing in the corresponding time sequence, wherein the associated knowledge tracking model is pre-trained according to a first loss and a second loss. The first loss is determined based on the prediction information of the mastery level of the specified knowledge point based on the training sample and its training label, and the second loss is determined based on the prediction information of the mastery level of the associated knowledge point set of the specified knowledge point based on the training sample and its training label.

[0055] In the embodiments of this specification, the associated knowledge tracking model can be pre-trained based on the total loss, which is determined by the weighted sum of the first loss and the second loss. Generally, the weight of the first loss can be set relatively higher to give more weight to the knowledge point being answered. For example, in the weighted sum, the weight ratio between the first loss and the second loss can be set to no less than 7:3.

[0056] The predicted information about the mastery level of a specific knowledge point can be used to predict the correctness of the answer to the specific knowledge point at the next moment.

[0057] To determine the first loss, first obtain all training samples with training labels within the sample time range. The training label corresponding to any time is the correct or incorrect answer information corresponding to the next time. Then, based on the deviation between the correct or incorrect answer information for the knowledge point answered at the next time based on the training sample and its corresponding training label, the single-sample loss of the training sample is determined. Finally, the first loss is determined based on the single-sample loss of the training sample.

[0058] The mastery level prediction information of the set of knowledge points associated with the specified knowledge point may be the correctness prediction information of the answer to the set of knowledge points associated with the specified knowledge point at the next moment.

[0059] When determining the second loss, the loss of a single associated knowledge point for the training sample can be determined based on the difference between the correct or incorrect answer prediction information for the set of associated knowledge points for the knowledge point to be answered at the next moment based on the training sample and the corresponding training labels. The second loss can then be determined based on the loss of the single associated knowledge point for the training sample.

[0060] If the user answers the knowledge point correctly or incorrectly, it can be predicted that the user should also answer the knowledge points in the associated knowledge point set correctly or incorrectly. In other words, if the user has mastered the knowledge point, then the user will also have a good grasp of the knowledge points in the associated knowledge point set related to the knowledge point.

[0061] In view of the above ideas, before determining the loss of a single associated knowledge point of a training sample, the embodiment of this specification can determine the training labels corresponding to the correct or incorrect answer prediction information of the associated knowledge point set to be consistent with the training labels corresponding to the correct or incorrect answer prediction information of the answered knowledge point. It should be noted that in this case, the training labels corresponding to the correct or incorrect answer prediction information of the associated knowledge point set (based on the labels corresponding to the answered knowledge points, which are forcibly corrected, and the correction can be temporary, and it can be considered to be modified back when the associated knowledge points become the answered knowledge points) may not conform to the actual situation of some users, but it helps to more fully learn the true correlation between knowledge points, so this processing method is still reasonable.

[0062] In the embodiment of this specification, when determining the first loss based on the single sample loss of the training sample, the average loss of the training sample can be determined based on the single sample loss of each training sample, and then the first loss can be determined based on the average loss of the training samples.

[0063] When determining the second loss based on the single-association knowledge point loss of the training sample, the embodiment of this specification can determine the average loss of the single-association knowledge point based on the single-association knowledge point loss of all training samples, and then determine the second loss based on the average loss of the single-association knowledge point.

[0064] S106 , predicting the user's mastery of the knowledge points in the predetermined knowledge point set after the specified time range through processing of the associated knowledge tracking model.

[0065] The embodiments of this specification can encode the question-answering interaction data when processed by the associated knowledge tracking model; and process the encoded results in turn through the deep neural network layer and the fully connected layer; then, the processing results output by the fully connected layer are mapped into probability information to represent the probability of the user answering correctly or incorrectly to the knowledge points in the predetermined knowledge point set after the specified time range, that is, the probability of answering the knowledge points correctly or incorrectly and the probability of answering correctly or incorrectly to the associated knowledge point set.

[0066] Since the question-answering interaction data includes the answered knowledge points, correct or incorrect answer information, and a set of associated knowledge points composed of the associated knowledge points of the answered knowledge points, the encoding of the question-answering interaction data can be to encode the answered knowledge points and their associated knowledge point sets separately according to the correct or incorrect answer information.

[0067] Therefore, when processing the associated knowledge tracking model, the answered knowledge points and their associated knowledge point sets can be encoded separately according to the correct or incorrect answer information. Through the encoding process, the first code corresponding to the answered knowledge point and the second code corresponding to the associated knowledge point set can be obtained.

[0068] Furthermore, the first code and the second code may be concatenated to serve as the encoding result of the corresponding question-answering interaction data.

[0069] The encoding results of the question-answering interaction data can be processed by a deep neural network layer and a fully connected layer; then, the processing results output by the fully connected layer are mapped into probability information to represent the probability of the user answering correctly or incorrectly to a knowledge point in a predetermined knowledge point set after a specified time range, that is, the probability of answering the knowledge point correctly or incorrectly and the probability of answering correctly or incorrectly to the associated knowledge point set.

[0070] Furthermore, the correct or incorrect answer information in the embodiments of this specification at least indicates whether the knowledge point answered is correct or incorrect. For the convenience of subsequent explanation, the first code is set to a 2M-dimensional unique hot vector, and the second code is set to an M'-dimensional vector, where M is the total number of knowledge points in the predetermined knowledge point set, and M=M'.

[0071] Based on the above settings, when encoding the answered knowledge point and its associated knowledge point set based on the correct or incorrect answer information, the embodiments of this specification can first determine the first M dimensions and the last M dimensions in the 2M dimensions. Then, based on the correct or incorrect answer indicated by the correct or incorrect answer information, determine whether to assign a value of 1 to the first M dimensions or the last M dimensions. Based on the sequence number of the answered knowledge point, the corresponding dimension is assigned a value of 1, and the other dimensions are assigned a value of 0, thereby obtaining a first encoding. Then, based on the sequence number of each associated knowledge point in the associated knowledge point set, the corresponding dimensions in the M' dimension are assigned a value of 1, and the other dimensions are assigned a value of 0, thereby obtaining a second encoding.

[0072] Furthermore, in the implementation process of the embodiments of this specification, see Figure 2 The schematic diagram of the knowledge points answered and their associated knowledge points is shown. The knowledge points answered in the figure are q t and q t The set of associated knowledge points S t , KC1 represents the knowledge points in the answered knowledge points, KC2 and KC3 represent the knowledge points in the associated knowledge point set, and the knowledge points can be obtained by processing the answered knowledge points q and the question p in the data set D. The question number is used as the key in the knowledge tracking data set, and the dictionary (Dic) of the question knowledge point set is combined with the question knowledge point set as the value. Let S be the associated knowledge point set of the knowledge point q of the answered question p, that is, S = {x|x∈Dic p ,x≠q}.

[0073] The associated knowledge tracking model can be summarized as follows: the user's interaction with the question at time t can be represented by the tuple h′ t =(q t ,S t, a t ) means, a tAnswer right or wrong information. Assume that the user's answer status at time t is H′ t ={h′0,……h′ t}, which can be used to predict the user’s understanding of knowledge point q at time t+1 t+1 Probability P(q t+1 )=P(a t+1 =1|q t+1 , H′ t ), or the user's overall knowledge mastery [P(q1), ..., P(q M )], the knowledge points answered t and q t The set of associated knowledge points S t .

[0074] The flow chart for generating the dictionary of question knowledge points can be found in Figure 3 First, input the data set, then obtain all the question numbers, then obtain the knowledge point set corresponding to the question number, establish the question-knowledge point set dictionary, and then return the dictionary to end the process. The specific steps are:

[0075] Process 1: First, traverse the knowledge tracking dataset to obtain the sequence number of the questions in the dataset.

[0076] Process 2: Then for each question number, traverse the entire data set to obtain a set of knowledge points with the same question number.

[0077] Process 3: Match each question number with the corresponding knowledge point set, and finally obtain the question knowledge point set dictionary for knowledge tracking through this process.

[0078] The dictionary of question knowledge points can be obtained through the following procedure:

[0079] Input:L p ,L q ,D(q,p),Dic

[0080] *L p ,L q Represents the question list and knowledge point list respectively; Dic represents the question-knowledge point set dictionary, D represents the data set composed of the answered knowledge point q and the question p, |D| is the number of data records in the data set D

[0081]

[0082]

[0083] For the associated knowledge tracking model, see Figure 4 The overall structure diagram of the associated knowledge tracking model is shown below, which is further explained in conjunction with the specific model structure.

[0084] Figure 4 The associated knowledge tracking model shown in the figure consists of an encoding layer, a deep neural network layer, a fully connected layer (FCL) and a prediction output layer (y t ).

[0085] Regarding the coding layer, it encodes the interactive data of answering questions. Specifically, it encodes the answered knowledge points and their related knowledge point sets respectively. Combined with the above-mentioned coding, the following is further explained through specific formulas.

[0086] For the encoding of the knowledge points answered, that is, encoding corresponding to the first encoding. For the knowledge points answered by the user, one-hot encoding with an encoding length of 2M can be performed. δ represents one-hot encoding. Suppose the knowledge point tuple answered by the user is (q, a), and the encoding matrix of the knowledge point answered is:

[0087] The specific formula for the first encoding is: Among them, q is the knowledge point answered, a is whether the knowledge point is answered correctly, which is 1 when the answer is correct, and M is the total number of knowledge points in the predetermined knowledge point set.

[0088] The encoding of the associated knowledge point, that is, the encoding corresponding to the second encoding, has a coding length of M. The associated knowledge point encoding encodes the vector of the knowledge point q and the associated knowledge point set S:

[0089] The specific formula for the second encoding is:

[0090] The deep neural network layer can adopt a common neural network. For example, the deep neural network layer is constructed using an RNN network.

[0091] For the output of the model, the Sigmoid activation function can be used to map the model output to the 0–1 interval, where the Sigmoid activation function is: It represents the model's prediction of the user's mastery of each knowledge point. The larger the value, the better the model predicts the user's mastery of the knowledge point. Let Logits be the output of the model's last fully connected layer (FCL). After the Sigmoid activation function, the model outputs the predicted value of the user's mastery of each knowledge point at time step t: t =Sigmoid(Logits)∈R M .

[0092] The loss function of the associated knowledge tracking model mentioned above is described in detail below:

[0093] The loss function of the model consists of two parts: the first loss representing the loss of the knowledge point answered And the second loss representing the loss of associated knowledge points Assume that the first loss and the second loss are obtained based on the cross entropy loss function. Of course, in addition to the cross entropy method, other methods such as mean square error and divergence can also be used to measure the loss.

[0094] in, Represents the cross entropy loss function, and the loss function is specifically: The embodiment of this specification can also train the associated knowledge tracking model through other loss functions, which is not limited to this.

[0095] In order to better determine the proportion of the loss of the answered knowledge points and the loss of the associated knowledge points, that is, to make the weight ratio λ between the first loss and the second loss play a greater role, the average answering interaction loss can be used for optimization to obtain the average loss function of the answered knowledge points and the average loss function of the associated knowledge points respectively.

[0096] Average loss function of the answered knowledge points:

[0097] Among them, δ(q t+1 ) represents the knowledge point q answered at the next moment t+1 t+1 One-hot encoding of , |H| represents the total number of user's question-answering interactions.

[0098] Average loss function of associated knowledge points:

[0099] Among them, q i ∈S t+1 , c i Indicates the correct or incorrect answer information of the related knowledge points. i , suppose the user's interactive situation in answering questions at time t can be represented by the tuple h′ t =(q t ,S t ,a t ) means, set c i =a t , this situation can be explained, once the user correctly or incorrectly answers q t , it can be predicted that the user should also answer the associated knowledge point set S correctly or incorrectly t In other words, if the user has mastered the knowledge point q t , then the user will have a good grasp of the knowledge points q t Related knowledge point set S t Related knowledge points in .

[0100] Total loss function express:

[0101]

[0102] λ is the weight ratio mentioned above, and λ can be set to be greater than or equal to 0.7. The above model loss function can optimize the loss of the knowledge point and the loss of the associated knowledge point at each time step, and the loss ratio of the associated knowledge point is determined by the weight coefficient λ. Set the loss function optimizer to Optimizer, and Θ to be the function Minimum value, the optimization objective function is:

[0103]

[0104] When applying the associated knowledge tracking model to the online tutoring system, the embodiments of this specification are mainly aimed at the knowledge point level online tutoring system. The following is a comparative explanation of the question level and the knowledge point level:

[0105] On the problem-level online tutoring system, see Figure 5 The schematic diagram of the question-level online tutoring system is shown in FIG. The user directly answers the question. If the question is answered correctly or incorrectly, all the knowledge points contained in the question are also answered correctly or incorrectly. Therefore, if the user answers the knowledge point q correctly or incorrectly, t , then he must answer the associated knowledge point set S correctly or incorrectly t Related knowledge points in. Because q t and S t From the same question. Here, a means the answer to the question is correct.

[0106] On the knowledge point level online tutoring system, see Figure 6 The schematic diagram of the knowledge point-level online tutoring system shown in the figure is much more complex than the question-level online tutoring system. The user can answer a knowledge point in the question individually and can answer the knowledge point once or multiple times. Therefore, if a participating user correctly answers knowledge point KC1 (the answered knowledge point), it does not mean that the participating user must correctly answer KC2 (the associated knowledge point). Among them, a1 indicates that the answer to knowledge point KC1 is correct, a2 indicates that the answer to knowledge point KC2 is correct, and a3 indicates that the answer to knowledge point KC3 is correct.

[0107] See Table 1 below, which shows the correct or incorrect answers for the knowledge point answered and its associated knowledge points at different times. s11 is the knowledge point answered, and s21 is its associated knowledge point. It can be seen that the user incorrectly answered s11 multiple times. Even if the user correctly answered s11 the fourth time, this indicates that the user's understanding of the knowledge point s11 is not good. Similarly, the user's understanding of the associated knowledge point s21 is also poor, which is why the user incorrectly answered s21 at the fifth time.

[0108] time Knowledge Points True or False Answer Information 1 s11 0 2 s11 0 3 s11 0 4 s11 1 5 s21 0

[0109] Table 1

[0110] On the surface, the knowledge points q t and the associated knowledge point set S t There is no obvious relationship between the correct answers of the knowledge points in the knowledge point level online tutoring system. However, there are a large number of user responses in the knowledge point level online tutoring system, as shown in Table 1, which shows that if the participating users answer the knowledge point q multiple times, t Error, even if the user eventually answers the knowledge point q correctly t , indicating that he has a good understanding of the knowledge point q t The user's mastery of the associated knowledge point set S is not good. t The knowledge points in the text are not well understood, and it is very likely that he will make mistakes in answering the related knowledge point set S. t Knowledge points in .

[0111] The embodiment of this specification also provides a prediction process of the associated knowledge tracking model. Figure 7 The following diagram shows an example of the prediction process of the associated knowledge tracking model in the open source dataset, and Table 2 shows the corresponding Figure 7 The knowledge point index and the user's true answer accuracy rate table for each knowledge point.

[0112]

[0113] Table 2

[0114] The dynamic process of the associated knowledge tracking model predicting the degree of knowledge mastery of users in the process of answering questions is as follows: Figure 7 The horizontal axis of the subgraph below is the knowledge points answered by the user and whether the answer is correct (q t ,a t ), the vertical axis of the lower subgraph is the index of the knowledge point answered, and the horizontal axis of the upper subgraph is the set of user-associated knowledge points {S t}, the vertical axis of the above subgraph is the index of the associated knowledge point. In Table 2, in the first three time steps t1-t3, s32 was answered correctly, and the model's prediction value for s32 began to increase. Because the associated knowledge point of s32 is s33, it can be seen that the model's prediction value for s33 is also gradually improving. Similarly, from the 4th time step to the 6th time step t4-t6, s33 was answered correctly, and the model's prediction value for s33 began to increase. At this time, the associated knowledge point of s33 is s32, so the prediction accuracy of s32 is also improving. From the 7th time step to the 9th time step t7-t9, s33 was answered correctly continuously, and the model's prediction probability for s33 began to rise. Because this s33 is a single knowledge point and has no associated knowledge points, the model only improved the prediction probability of s33. At the last time step, t10, the user correctly answers question s37. The prediction accuracy for both s37 and its associated knowledge point, s55, also improves. This shows that the prediction process of the associated knowledge tracking model closely matches the user's actual knowledge level.

[0115] It can be seen that the model of the embodiment of this specification has better prediction accuracy and prediction breadth, and can better simulate the user's overall knowledge system at each moment.

[0116] Figure 8 A schematic structural diagram of an associated knowledge tracking device provided for one or more embodiments of this specification includes: an acquisition unit 802 , an input unit 804 and a prediction unit 806 .

[0117] The acquisition unit 802 acquires the user's answer interaction data within a specified time range, the answer interaction data including the knowledge point answered at the corresponding time, the correct or incorrect answer information, and the associated knowledge point set consisting of the associated knowledge points of the answered knowledge point;

[0118] The input unit 804 inputs the interactive data of each question answering into the associated knowledge tracking model in the corresponding time sequence for processing, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on the mastery level prediction information of the specified knowledge point based on the training sample and its training label, and the second loss is determined based on the mastery level prediction information of the set of associated knowledge points of the specified knowledge point based on the training sample and its training label;

[0119] The prediction unit 806 predicts the user's mastery of the knowledge points in the predetermined knowledge point set after a specified time range by processing the associated knowledge tracking model.

[0120] Furthermore, the related knowledge points of the answered knowledge points include:

[0121] One or more knowledge points other than the knowledge points answered in the question.

[0122] Furthermore, when the prediction unit 806 executes the associated knowledge tracking model processing, it specifically includes:

[0123] Encode the question-answering interaction data;

[0124] The encoding results are processed in sequence through the deep neural network layer and the fully connected layer;

[0125] The processing results output by the fully connected layer are mapped into probability information to represent the probability of the user answering correctly or incorrectly to a knowledge point in the predetermined knowledge point set after a specified time range.

[0126] Furthermore, when the prediction unit 806 executes the associated knowledge tracking model processing, it specifically includes:

[0127] According to the correct or incorrect answer information, the knowledge points answered and their related knowledge points are coded respectively;

[0128] Through the coding process, a first code corresponding to the answered knowledge point and a second code corresponding to the associated knowledge point set are obtained.

[0129] Furthermore, the prediction unit 806 performs the processing of the associated knowledge tracking model, further comprising:

[0130] The first code and the second code are concatenated to obtain the encoding result of the corresponding question-answering interaction data.

[0131] Furthermore, the correct or incorrect answer information at least indicates whether the knowledge point answered is correct or incorrect. The first encoding is a 2M-dimensional one-hot vector, and the second encoding is an M'-dimensional vector, where M is the total number of knowledge points in the predetermined knowledge point set, M=M';

[0132] When the prediction unit 806 performs encoding of the answered knowledge point and its associated knowledge point set according to the correct or incorrect answer information, it specifically includes:

[0133] Determine the first M dimensions and the last M dimensions in the 2M dimensions;

[0134] According to the correct or incorrect answer information, determine whether to assign a value of 1 to the first M dimensions or the last M dimensions, and according to the sequence number of the knowledge point answered, assign a value of 1 to the corresponding dimension and assign a value of 0 to the other dimensions, thereby obtaining a first code;

[0135] According to the sequence number of each associated knowledge point in the associated knowledge point set, the corresponding dimensions in the M′ dimension are assigned a value of 1, and the other dimensions are assigned a value of 0, thereby obtaining a second code.

[0136] Furthermore, before the acquisition unit 806 acquires the user's interactive data on each question within a specified time range, the device further includes:

[0137] The independent unit reads the original data set of questions and knowledge points from the relational database;

[0138] Traverse the unit, traverse the original data set, and obtain the sequence number of each question;

[0139] Get the unit, traverse the original data set according to the serial number of the question, and get the set of knowledge points whose serial numbers are consistent with the serial numbers of the question;

[0140] The construction unit constructs a question knowledge point set dictionary according to the correspondence between the question number and the set of knowledge points, so as to obtain the answered knowledge points and the associated knowledge point sets of the answered questions by querying the question knowledge point set dictionary.

[0141] Furthermore, the associated knowledge tracking model is pre-trained based on the total loss, which is determined by the weighted sum of the first loss and the second loss;

[0142] Among them, in the weighted sum, the weight ratio between the first loss and the second loss is not less than 7:3.

[0143] Furthermore, the first loss and the second loss are determined as follows:

[0144] Obtain each training sample with a training label within the sample time range, where the training label corresponding to any time is the correct or incorrect answer information corresponding to the next time;

[0145] Determine the single sample loss of the training sample based on the prediction of the correct or incorrect answer to the knowledge point to be answered at the next moment based on the training sample and the degree of deviation between the corresponding training label;

[0146] Determine the first loss based on the single sample loss of the training sample;

[0147] Determine the single associated knowledge point loss of the training sample based on the correct or incorrect answer prediction information of the associated knowledge point set of the knowledge point to be answered at the next moment based on the training sample and the degree of difference between the corresponding training labels;

[0148] The second loss is determined based on the single associated knowledge point loss of the training sample.

[0149] Furthermore, when the input unit 804 performs the single sample loss based on the training sample to determine the first loss, it specifically includes:

[0150] According to the single sample loss of each training sample, determine the average loss of the training samples;

[0151] Determine a first loss based on the average loss of the training samples; and / or,

[0152] The second loss is determined based on the single associated knowledge point loss of the training sample, specifically including:

[0153] Determine the average loss of single-association knowledge points based on the single-association knowledge point losses of all training samples;

[0154] The second loss is determined based on the average loss of single associated knowledge points.

[0155] Furthermore, before the input unit 804 determines the single associated knowledge point loss of the training sample, the apparatus further includes:

[0156] The determination unit determines that the training label corresponding to the correctness prediction information of the associated knowledge point set is consistent with the training label corresponding to the correctness prediction information of the answered knowledge point.

[0157] Figure 9 A schematic diagram of a related knowledge tracking device is provided for one or more embodiments of this specification. The device is applied to a third-party platform application. The mobile terminal operation data is held by the communication operator used by the user, including:

[0158] at least one processor; and,

[0159] a memory communicatively connected to at least one processor; wherein,

[0160] The memory stores instructions executable by at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0161] Obtaining the user's interactive data on each question within a specified time range, the interactive data including the knowledge points answered at the corresponding time, correct or incorrect answer information, and a set of associated knowledge points consisting of associated knowledge points of the knowledge points answered;

[0162] Input each question-answering interaction data into the associated knowledge tracking model in the corresponding time sequence for processing, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on the mastery prediction information of the specified knowledge point based on the training samples and its training labels, and the second loss is determined based on the mastery prediction information of the set of associated knowledge points of the specified knowledge point based on the training samples and its training labels;

[0163] Through the processing of the associated knowledge tracking model, the user's mastery of the knowledge points in the predetermined knowledge point set after the specified time range is predicted.

[0164] One or more embodiments of this specification provide a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured to:

[0165] Obtaining the user's interactive data on each question within a specified time range, the interactive data including the knowledge points answered at the corresponding time, correct or incorrect answer information, and a set of associated knowledge points consisting of associated knowledge points of the knowledge points answered;

[0166] Input each question-answering interaction data into the associated knowledge tracking model in the corresponding time sequence for processing, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on the mastery prediction information of the specified knowledge point based on the training samples and its training labels, and the second loss is determined based on the mastery prediction information of the set of associated knowledge points of the specified knowledge point based on the training samples and its training labels;

[0167] Through the processing of the associated knowledge tracking model, the user's mastery of the knowledge points in the predetermined knowledge point set after the specified time range is predicted.

[0168] In the 1990s, technological improvements could be clearly distinguished as either hardware improvements (for example, improvements to circuit structures like diodes, transistors, and switches) or software improvements (improvements to process flows). However, with the advancement of technology, many process flow improvements today can now be considered direct improvements to hardware circuit structures. Designers almost always create the corresponding hardware circuit structure by programming the improved process flow into the hardware circuit. Therefore, it cannot be said that a process flow improvement cannot be implemented using hardware modules. For example, a programmable logic device (PLD), such as a field programmable gate array (FPGA), is an integrated circuit whose logical function is determined by user programming. Designers can "integrate" a digital system on a PLD through their own programming, without having to hire a chip manufacturer to design and manufacture a dedicated integrated circuit chip. Moreover, nowadays, instead of manually fabricating integrated circuit chips, this programming is mostly done using "logic compiler" software. This is similar to the software compiler used when developing programs. Before compilation, the original code must also be written in a specific programming language, called a hardware description language (HDL). There is not just one HDL, but many, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art will also understand that by simply programming the method flow in one of these hardware description languages ​​and then programming it into an integrated circuit, a hardware circuit that implements the logic method flow can be easily obtained.

[0169] The controller can be implemented in any suitable manner. For example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to implementing the controller in a purely computer-readable program code format, the controller can be implemented in the form of logic gates, switches, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller can be considered a hardware component, and the means for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the means for implementing various functions can be considered as both a software module implementing the method and a structure within the hardware component.

[0170] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0171] For the convenience of description, the above devices are described as being divided into various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0172] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0173] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0174] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0175] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0176] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0177] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0178] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules 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, compact disc read-only memory (CD-ROM), digital versatile disc (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. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0179] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0180] This specification may be described in the general context of computer-executable instructions, such as program modules, executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media, including storage devices.

[0181] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simplified. For relevant details, refer to the descriptions of the method embodiments.

[0182] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0183] The above is merely one or more embodiments of this specification and is not intended to limit this specification. It will be apparent to those skilled in the art that various modifications and variations may be made to one or more embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of one or more embodiments of this specification shall be included within the scope of the claims of this specification.

Claims

1. A method for tracking associated knowledge, comprising: Obtaining interactive data on each question answering by the user within a specified time range, the interactive data including the knowledge points answered at the corresponding time, correct or incorrect information about the answers, and a set of associated knowledge points consisting of associated knowledge points of the knowledge points answered; Inputting the various question-answering interaction data into an associated knowledge tracking model for processing in a corresponding time sequence, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on prediction information of the mastery level of a specified knowledge point based on a training sample and its training label, and the second loss is determined based on prediction information of the mastery level of a set of associated knowledge points of the specified knowledge point based on the training sample and its training label; Predicting the user's mastery of knowledge points in a predetermined knowledge point set after the specified time range through the processing of the associated knowledge tracking model; The first loss and the second loss are determined as follows: Obtaining training samples with training labels within a sample time range, wherein the training label corresponding to any time is the correct or incorrect answer information corresponding to the next time; Determining the single sample loss of the training sample based on the prediction information of the correctness or incorrectness of the answer to the knowledge point to be answered at the next moment based on the training sample and the degree of deviation between the corresponding training label; Determining the average loss of the training samples according to the single sample loss of each training sample; determining the first loss according to the average loss of the training samples; Determining the single associated knowledge point loss of the training sample based on the prediction information of the correct or incorrect answer of the associated knowledge point set of the knowledge point to be answered at the next moment based on the training sample and the degree of difference between the corresponding training labels; According to the single-association knowledge point losses of all the training samples, the average loss of the single-association knowledge points is determined; and according to the average loss of the single-association knowledge points, the second loss is determined.

2. The method according to claim 1, wherein the related knowledge points of the answered knowledge point include: One or more knowledge points in the answered question other than the answered knowledge point.

3. The method according to claim 1, wherein the processing of the associated knowledge tracking model specifically comprises: Encoding the question-answering interaction data; The encoding results are processed in sequence through the deep neural network layer and the fully connected layer; The processing result output by the fully connected layer is mapped into probability information to represent the probability of the user answering correctly or incorrectly to a knowledge point in a predetermined knowledge point set after the specified time range.

4. The method according to claim 1, wherein the processing of the associated knowledge tracking model specifically comprises: According to the correctness information of the answer, the answered knowledge point and the associated knowledge point set are respectively encoded; Through the encoding process, a first code corresponding to the answered knowledge point and a second code corresponding to the associated knowledge point set are obtained.

5. The method of claim 4, wherein the processing of the associated knowledge tracking model further comprises: The first code and the second code are concatenated as the encoding result of the corresponding question-answering interaction data.

6. The method of claim 4, wherein the correct / incorrect answer information at least indicates whether the knowledge point answered is correct or incorrect, the first encoding is a 2M-dimensional one-hot vector, and the second encoding is an M'-dimensional vector, where M is the total number of knowledge points in the predetermined knowledge point set, and M=M'; The step of encoding the answered knowledge point and its associated knowledge point set according to the correct or incorrect answer information specifically includes: Determine the first M dimensions and the last M dimensions in the 2M dimensions; Determining whether to assign a value of 1 to the first M dimensions or the last M dimensions based on whether the answer is correct or incorrect as indicated by the correct or incorrect answer information, and assigning a value of 1 to the corresponding dimension and 0 to the other dimensions based on the sequence number of the knowledge point answered, thereby obtaining the first code; According to the serial number of each associated knowledge point in the associated knowledge point set, the corresponding dimensions in the M′ dimension are assigned a value of 1, and the other dimensions are assigned a value of 0, thereby obtaining the second code.

7. The method according to claim 1, before obtaining the user's interactive data on each question within a specified time range, the method further comprises: Read the original dataset of questions and knowledge points from the relational database; Traversing the original data set to obtain the sequence number of each question; According to the serial number of the question, the original data set is traversed to obtain a set of knowledge points whose serial numbers are consistent with the serial number of the question; According to the correspondence between the serial number of the question and the set of the knowledge points, a question knowledge point set dictionary is constructed so that the answered knowledge points of the answered question and their associated knowledge point sets can be obtained by querying the question knowledge point set dictionary.

8. The method of claim 1 , wherein the associated knowledge tracking model is pre-trained based on a total loss, and the total loss is determined based on a weighted sum of the first loss and the second loss; in, In the weighted sum, a weight ratio between the first loss and the second loss is not less than 7:

3.

9. The method according to claim 1, before determining the single-association knowledge point loss of the training sample, the method further comprises: The training labels corresponding to the correct / wrong answer prediction information of the associated knowledge point set are determined to be consistent with the training labels corresponding to the correct / wrong answer prediction information of the answered knowledge point.

10. A device for tracking associated knowledge, comprising: an acquisition unit, which acquires each question-answering interaction data of the user within a specified time range, wherein the question-answering interaction data includes the knowledge point answered at the corresponding time, the correct or incorrect answer information, and a set of associated knowledge points consisting of associated knowledge points of the knowledge point answered; an input unit that inputs the question-answering interaction data into an associated knowledge tracking model for processing in a corresponding time sequence, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on prediction information of a mastery level of a specified knowledge point based on a training sample and its training label, and the second loss is determined based on prediction information of a mastery level of a set of associated knowledge points of the specified knowledge point based on the training sample and its training label; a prediction unit, configured to predict the user's mastery of knowledge points in a predetermined knowledge point set after the specified time range through the processing of the associated knowledge tracking model; The first loss and the second loss are determined as follows: Obtaining training samples with training labels within a sample time range, wherein the training label corresponding to any time is the correct or incorrect answer information corresponding to the next time; Determining the single sample loss of the training sample based on the prediction information of the correctness or incorrectness of the answer to the knowledge point to be answered at the next moment based on the training sample and the degree of deviation between the corresponding training label; Determining the average loss of the training samples according to the single sample loss of each training sample; determining the first loss according to the average loss of the training samples; Determining the single associated knowledge point loss of the training sample based on the prediction information of the correct or incorrect answer of the associated knowledge point set of the knowledge point to be answered at the next moment based on the training sample and the degree of difference between the corresponding training labels; According to the single-association knowledge point losses of all the training samples, the average loss of the single-association knowledge points is determined; and according to the average loss of the single-association knowledge points, the second loss is determined.

11. The apparatus according to claim 10, wherein the related knowledge points of the answered knowledge point include: One or more knowledge points in the answered question other than the answered knowledge point.

12. The apparatus according to claim 10, wherein the prediction unit, when executing the processing of the associated knowledge tracking model, specifically comprises: Encoding the question-answering interaction data; The encoding results are processed in sequence through the deep neural network layer and the fully connected layer; The processing result output by the fully connected layer is mapped into probability information to represent the probability of the user answering correctly or incorrectly to a knowledge point in a predetermined knowledge point set after the specified time range.

13. The apparatus according to claim 10, wherein the prediction unit, when executing the processing of the associated knowledge tracking model, specifically comprises: According to the correctness information of the answer, the answered knowledge point and the associated knowledge point set are respectively encoded; Through the encoding process, a first code corresponding to the answered knowledge point and a second code corresponding to the associated knowledge point set are obtained.

14. The apparatus according to claim 13, wherein the prediction unit performs the processing of the associated knowledge tracking model, further comprising: The first code and the second code are concatenated as the encoding result of the corresponding question-answering interaction data.

15. The apparatus of claim 13, wherein the correct / incorrect answer information at least indicates whether the knowledge point answered is correct or incorrect, the first code is a 2M-dimensional one-hot vector, and the second code is an M'-dimensional vector, where M is the total number of knowledge points in the predetermined knowledge point set, and M=M'; When the prediction unit performs the encoding of the answered knowledge point and the associated knowledge point set according to the correct or incorrect answer information, the method specifically includes: Determine the first M dimensions and the last M dimensions in the 2M dimensions; Determining whether to assign a value of 1 to the first M dimensions or the last M dimensions based on whether the answer is correct or incorrect as indicated by the correct or incorrect answer information, and assigning a value of 1 to the corresponding dimension and 0 to the other dimensions based on the sequence number of the knowledge point answered, thereby obtaining the first code; According to the serial number of each associated knowledge point in the associated knowledge point set, the corresponding dimensions in the M′ dimension are assigned a value of 1, and the other dimensions are assigned a value of 0, thereby obtaining the second code.

16. The device according to claim 10, wherein before the acquiring unit executes the step of acquiring the interactive data of each question answering by the user within the specified time range, the device further comprises: The independent unit reads the original data set of questions and knowledge points from the relational database; A traversal unit traverses the original data set to obtain a sequence number of each question; An obtaining unit, for the serial number of the question, traversing the original data set to obtain a set of knowledge points whose serial numbers are consistent with the serial number of the question; The construction unit constructs a question knowledge point set dictionary according to the correspondence between the serial number of the question and the set of knowledge points, so as to obtain the answered knowledge points and the associated knowledge point sets of the answered question by querying the question knowledge point set dictionary.

17. The apparatus of claim 10, wherein the associated knowledge tracking model is pre-trained according to a total loss, and the total loss is determined according to a weighted sum of the first loss and the second loss; in, In the weighted sum, a weight ratio between the first loss and the second loss is not less than 7:

3.

18. The apparatus according to claim 10, wherein before the input unit determines the single-association knowledge point loss of the training sample, the apparatus further comprises: The determination unit determines that the training label corresponding to the correctness prediction information of the answer of the associated knowledge point set is consistent with the training label corresponding to the correctness prediction information of the answer of the knowledge point.

19. A device for tracking associated knowledge, comprising: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Obtaining interactive data on each question answering by the user within a specified time range, the interactive data including the knowledge points answered at the corresponding time, correct or incorrect information about the answers, and a set of associated knowledge points consisting of associated knowledge points of the knowledge points answered; Inputting the various question-answering interaction data into an associated knowledge tracking model for processing in a corresponding time sequence, wherein the associated knowledge tracking model is pre-trained based on a first loss and a second loss, wherein the first loss is determined based on prediction information of the mastery level of a specified knowledge point based on a training sample and its training label, and the second loss is determined based on prediction information of the mastery level of a set of associated knowledge points of the specified knowledge point based on the training sample and its training label; Predicting the user's mastery of knowledge points in a predetermined knowledge point set after the specified time range through the processing of the associated knowledge tracking model; The first loss and the second loss are determined as follows: Obtaining training samples with training labels within a sample time range, wherein the training label corresponding to any time is the correct or incorrect answer information corresponding to the next time; Determining the single sample loss of the training sample based on the prediction information of the correctness or incorrectness of the answer to the knowledge point to be answered at the next moment based on the training sample and the degree of deviation between the corresponding training label; Determining the average loss of the training samples according to the single sample loss of each training sample; determining the first loss according to the average loss of the training samples; Determining the single associated knowledge point loss of the training sample based on the prediction information of the correct or incorrect answer of the associated knowledge point set of the knowledge point to be answered at the next moment based on the training sample and the degree of difference between the corresponding training labels; According to the single-association knowledge point losses of all the training samples, the average loss of the single-association knowledge points is determined; and according to the average loss of the single-association knowledge points, the second loss is determined.

Images