A teaching robot teaching abnormality recognition prompt method and system

Abstract

The present invention discloses a teaching abnormality identification and prompting method and system for a teaching robot, comprising: obtaining a real-time learning behavior vector of a trainee to be identified within a set time window; inputting the real-time learning behavior vector into a pre-trained abnormality level recognition model, and outputting an abnormality level label corresponding to the real-time learning behavior vector; wherein, the training of the abnormality level recognition model is based on the sorting deviation and reconstruction deviation of the historical behavior sequence to fit the evolution of the learning behavior vector in the historical data; triggering the teaching abnormality prompt of the teaching robot according to the abnormality level label; the present invention not only effectively gets rid of the dependence on artificial experience and labels, but also significantly enhances the ability to characterize the behavior evolution trajectory; it not only measures the overall reconstruction deviation of the behavior sequence in the feature space, but also further evaluates its sorting deviation in the temporal structure, thereby improving the pertinence of the abnormality recognition prompt.

Description

Technical Field

[0001] The present invention relates to the field of teaching abnormality recognition and prompting, and in particular to a teaching abnormality recognition and prompting method and system for a teaching robot. Background Art

[0002] With the in-depth application of artificial intelligence in education, teaching robots are gradually taking on the task of dynamically identifying students' learning status and providing personalized intervention. In particular, in intelligent teaching systems based on big data, how to use behavioral data to identify students' abnormal learning behaviors in real time has become a key issue in improving learning outcomes.

[0003] Currently, common anomaly recognition methods are mostly based on rule extraction or traditional classification models, which rely mainly on manually set threshold conditions, static behavioral characteristics, or pre-labeled anomaly labels for training. These methods have the following problems:

[0004] First, the construction of supervised labels relies on manual annotation or static rules, which are highly subjective and have poor scalability. Because learning anomalies in teaching scenarios are often hidden in temporal behavioral changes, it is difficult for humans to accurately identify the degree of anomalies at different levels. This leads to strong dependence on model training and high label distortion, making it difficult to adapt to the classification needs of large-scale data.

[0005] Second, existing models struggle to fully capture the temporal evolution of behavioral sequences. In actual instruction, students' learning behaviors exhibit not only spatial deviations from standard behavior (i.e., inconsistent behavioral characteristics), but also structural differences in order (i.e., the order of behavioral changes). Existing technologies generally ignore modeling these ordering deviations, making it difficult for recognition models to capture the implicit evolutionary trajectory of learning behavior and the periodic characteristics of learning regression. Summary of the Invention

[0006] In view of the shortcomings of the existing technology, the present invention provides a teaching anomaly identification and prompting method and system for a teaching robot, which solves the technical problems raised in the background technology by fitting the sorting deviation and reconstruction deviation in historical data.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0008] A teaching abnormality recognition and prompting method and system for a teaching robot, comprising the following steps:

[0009] Obtain the real-time learning behavior vector of the student to be identified within the set time window;

[0010] Inputting the real-time learning behavior vector into a pre-trained abnormality level recognition model, and outputting an abnormality level label corresponding to the real-time learning behavior vector;

[0011] The training of the anomaly level recognition model is based on the sorting deviation and reconstruction deviation of the historical behavior sequence to fit the evolution of the learning behavior vector in the historical data;

[0012] According to the abnormality level label, the teaching robot's teaching abnormality prompt is triggered.

[0013] In some embodiments, the step of building the abnormality level identification model includes:

[0014] S1. Obtain the first-stage scores of N target students;

[0015] S2. Construct a historical behavior sequence based on the first-stage scores of N target students; the historical behavior sequence includes: K high-performance behavior sequences and Q low-performance behavior sequences;

[0016] S3. defining the K high-performance behavior sequences as an autoencoding training set;

[0017] S4. Iteratively train the temporal autoencoder model using the autoencoder training set to obtain a temporal behavior reconstruction model that can fit high-performance behavior sequences;

[0018] S5. Input Q low-performance behavior sequences into the temporal behavior reconstruction model and output Q reconstructed behavior sequences;

[0019] S6. Based on the Q reconstructed behavior sequences and the corresponding low-performance behavior sequences, select non-extremely abnormal students from the target students;

[0020] S7. Use high-performance behavior sequences and low-performance behavior sequences of non-extremely abnormal students to construct a self-supervised training set;

[0021] S8. Use the self-supervised training set to perform supervised training on the time series modeling model to obtain the abnormality level recognition model.

[0022] In some embodiments, the step of constructing K high-performance behavior sequences and Q low-performance behavior sequences includes:

[0023] S2-1. Compare the first-stage scores of the target students with the excellent score threshold and the passing score threshold respectively;

[0024] If the target student's first-stage score is higher than the excellent score threshold, the target student will be marked as a first-category student;

[0025] If the target student's first-stage score is lower than the passing score threshold, the target student will be marked as a second-category student;

[0026] S2-2, obtaining the second-stage scores and the second-stage learning behavior vectors of the first and second-stage students;

[0027] S2-3, compare the second stage scores with the excellent score threshold and the passing score threshold respectively;

[0028] If the second stage score is higher than the excellent score threshold, the first category student or the second category student to whom the second stage score belongs will be marked as a third category student;

[0029] If the second stage score is lower than the passing score threshold, the first category student or the second category student to which the second stage score belongs is marked as a fourth category student;

[0030] S2-4, from the learning behavior vectors of the first and second types of students, respectively select M learning behavior vectors marked as the third and fourth types of students;

[0031] S2-5, sorting the M learning behavior vectors of the third and fourth types of students respectively to generate corresponding high-performance behavior sequences and low-performance behavior sequences;

[0032] S2-6. Repeat S2-1 to S2-5 until K high-performance behavior sequences and Q low-performance behavior sequences are obtained.

[0033] In some embodiments, obtaining the second-stage scores and the second-stage learning behavior vectors of the first and second-stage students includes:

[0034] S2-2-1. In the second stage, a fixed-length and sliding timing window is defined;

[0035] S2-2-2, sliding the time series window with a set sliding step size, and sequentially obtaining the learning behavior vectors of the first and second types of students in the second stage within M time series windows;

[0036] S2-2-3. When the timing window slides to the end of the second stage, the second stage scores of the first and second types of students are collected.

[0037] In some embodiments, generating corresponding high-performance behavior sequences and low-performance behavior sequences includes:

[0038] S2-5-1. Obtain the central timestamps of M time series windows;

[0039] S2-5-2. Encode the central timestamps of the M time series windows into time series numbers;

[0040] S2-5-3. Sort the learning behavior vectors of the third and fourth categories of students respectively according to the timing numbers of the M timing windows to obtain the high-performance behavior sequence and the low-performance behavior sequence.

[0041] In some embodiments, screening non-extremely abnormal students from target students includes:

[0042] S6-1. Calculate the ranking deviations between the Q reconstructed behavior sequences and the corresponding low-performance behavior sequences to obtain Q ranking deviations for the fourth type of students;

[0043] The expression of the sorting deviation is:

[0044] ;

[0045] in, is the sorting deviation, M is the number of time windows, is the ranking rank of the low-performance behavior sequence in the t-th time window, Reconstruct the sorting rank in the behavior sequence within the t-th time window;

[0046] S6-2. From the Q ranking deviations, select J ranking deviations that are lower than the ranking deviation threshold;

[0047] S6-3. Mark the fourth category of students to which J ranking deviations belong as non-extreme abnormal students.

[0048] In some embodiments, constructing the self-supervised training set includes:

[0049] S7-1. Filter out J reconstructed behavior sequences of students marked as non-extremely abnormal;

[0050] S7-2. Calculate the reconstruction deviations between the J reconstructed behavior sequences and the corresponding low-performance behavior sequences to obtain J reconstruction deviations for non-extremely abnormal students;

[0051] The expression of the reconstruction deviation is:

[0052] ;

[0053] in, To reconstruct the deviation, is the learning behavior vector in the low-performance behavior sequence in the t-th time window, is the reconstructed behavior vector in the reconstructed behavior sequence within the t-th time window; is the square of the Euclidean distance between the learned behavior vector in the low-performance behavior sequence and the reconstructed behavior vector in the reconstructed behavior sequence in all feature dimensions;

[0054] S7-3. Sort the J reconstruction deviations by size, and assign J abnormality level labels to the low-performance behavior sequences of the non-extremely abnormal students in sequence; the abnormality level label of the smallest reconstruction deviation is set to 1;

[0055] S7-4, uniformly setting the abnormality level labels of the K high-performance behavior sequences of the third-category students to 0, to form corresponding K abnormality level labels;

[0056] S7-5. Pair J anomaly level labels with corresponding low-performance behavior sequences, and pair K anomaly level labels with corresponding high-performance behavior sequences to construct a self-supervised training set containing S anomaly supervision samples; where S = J + K.

[0057] Compared to existing technologies, the present invention's teaching robot teaching anomaly recognition and prompting method uses an autoencoder model trained on historically high-performance behaviors. Leveraging its ability to reconstruct sequences of low-performance behaviors, it automatically quantifies the degree of deviation between these sequences and ideal behaviors and generates multi-level anomaly labels based on these deviations. This method effectively eliminates reliance on manual experience and labels and exhibits superior self-supervised training capabilities.

[0058] Furthermore, this invention significantly enhances its ability to characterize behavioral evolutionary trajectories by introducing dual fitting metrics: ranking bias and reconstruction bias. This not only measures the overall reconstruction bias of behavioral sequences in feature space, but also further assesses ranking bias within the temporal structure. This allows for precise identification of non-extremely anomalous students with stable behavior but strategic deviations, thereby improving the targeted nature of anomaly identification.

[0059] In a second aspect, the present invention provides a teaching abnormality recognition and prompting system for a teaching robot, comprising:

[0060] Vector acquisition module, used to obtain the real-time learning behavior vector of the student to be identified within the set time window

[0061] An abnormality level output module, used to input the real-time learning behavior vector into a pre-trained abnormality level recognition model and output an abnormality level label corresponding to the real-time learning behavior vector;

[0062] The training of the abnormal level recognition model is based on the sorting deviation and reconstruction deviation of the historical behavior sequence to fit the evolution of the learning behavior vector in the historical data.

[0063] The prompt module is used to trigger the teaching robot's teaching exception prompt according to the exception level label.

[0064] Compared with the prior art, the beneficial effects of the teaching abnormality identification and prompting system of the teaching robot of the present invention are the same as the beneficial effects of the teaching abnormality identification and prompting method of the teaching robot described above, so they will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A schematic diagram of a flow chart of a teaching abnormality recognition and prompting method for a teaching robot according to the present invention;

[0066] Figure 2 Schematic diagram of the modeling process of the abnormal level identification model of the present invention;

[0067] Figure 3 This is a schematic diagram of the process for marking non-extremely abnormal students according to the present invention;

[0068] Figure 4 This is a structural block diagram of a teaching abnormality recognition and prompting system for a teaching robot according to the present invention. DETAILED DESCRIPTION

[0069] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0070] See also Figure 1 The present invention provides a teaching abnormality recognition and prompting method for a teaching robot, comprising the following steps:

[0071] A1. Obtain the real-time learning behavior vector of the student to be identified within a set time window;

[0072] The learning behavior vector is a combination of multi-dimensional learning behavior features collected within the current time window, and the feature dimensions include but are not limited to question response time, page dwell time, click frequency, error rate distribution, and behavior pause time.

[0073] A2. Input the real-time learning behavior vector into a pre-trained abnormality level recognition model, and output an abnormality level label corresponding to the real-time learning behavior vector;

[0074] The training of the anomaly level recognition model is based on the sorting deviation and reconstruction deviation of the historical behavior sequence to fit the evolution of the learning behavior vector in the historical data;

[0075] The anomaly level recognition model is a time-series classification model trained based on historical high- and low-performing student behavior samples. It outputs labels for behavior sequences at different anomaly levels. These labels are discrete values ​​representing the degree of deviation between a student's behavior and high-performing behavior. Larger values ​​indicate a greater degree of deviation. Examples include: Level 0 (normal), Level 1 (mildly abnormal), Level 2 (moderately abnormal), and Level 3 (severely abnormal).

[0076] A3. Trigger the teaching robot's teaching abnormality prompt based on the abnormality level label.

[0077] For example, when the identified abnormality level label is 1 to 3, the teaching robot generates personalized teaching intervention prompts corresponding to the identification level based on different level strategies, including learning strategy suggestions, behavioral reminders or teacher assistance requests, etc., to dynamically guide students to adjust their learning paths and improve learning outcomes.

[0078] See Figures 2 to 3 This embodiment also discloses a modeling step for an abnormality level recognition model, and the modeling step includes:

[0079] S1. Obtain the first-stage scores of N target students;

[0080] For example, the first stage score may be the target student's final score for the previous semester.

[0081] S2. Construct a historical behavior sequence based on the first-stage scores of N target students; the historical behavior sequence includes: K high-performance behavior sequences and Q low-performance behavior sequences;

[0082] S3. defining the K high-performance behavior sequences as an autoencoding training set;

[0083] S4. Iteratively train the temporal autoencoder model using the autoencoder training set to obtain a temporal behavior reconstruction model that can fit high-performance behavior sequences;

[0084] In this embodiment, the temporal autoencoder model may adopt a sequence modeling structure consisting of an encoder and a decoder. Specifically:

[0085] The encoder is used to receive and compress the time series features in the high-performance behavior sequence, and perform nonlinear mapping on the learning behavior vectors at different moments to extract their potential time series feature representations.

[0086] The decoder is used to reconstruct a behavior vector sequence with the same length as the input high-performance behavior sequence based on the temporal feature representation extracted by the encoder.

[0087] Both the encoder and decoder modules can employ stacked long-short-term memory (LSTM) or gated recurrent unit (GRU) structures to enhance the ability to model long-term dependencies within behavioral sequences. During model training, the high-performance behavioral sequences themselves serve as supervisory signals, and the model parameters are iteratively optimized by minimizing the reconstruction deviation between the input high-performance behavioral sequence and the reconstructed sequence. Ultimately, the trained temporal behavior reconstruction model can accurately reconstruct high-performance behavioral sequences.

[0088] S5. Input Q low-performance behavior sequences into the temporal behavior reconstruction model and output Q reconstructed behavior sequences;

[0089] S6. Based on the Q reconstructed behavior sequences and the corresponding low-performance behavior sequences, select non-extremely abnormal students from the target students;

[0090] S7. Use high-performance behavior sequences and low-performance behavior sequences of non-extremely abnormal students to construct a self-supervised training set;

[0091] S8. Use the self-supervised training set to perform supervised training on the time series modeling model to obtain the abnormality level recognition model.

[0092] Specifically, the time series modeling model can use a bidirectional long short-term memory (BiLSTM) structure to extract temporal dependencies within a student's behavioral sequence. A fully connected classification layer and a softmax activation function are set at the output to predict the abnormality level label for each behavioral sequence. The supervised training modeling steps can be similar to the common supervised training steps.

[0093] Exemplarily, the steps for obtaining the high-performance sequence and the low-performance sequence are:

[0094] S2-1. Compare the first-stage scores of the target students with the excellent score threshold and the passing score threshold respectively;

[0095] If the target student's first-stage score is higher than the excellent score threshold, the target student will be marked as a first-category student, which is used to represent the high-performing student group;

[0096] If the target student's first-stage score is lower than the passing score threshold, the target student will be marked as a second-category student; used to represent the low-performing student group.

[0097] S2-2, obtaining the second-stage scores and the second-stage learning behavior vectors of the first and second-stage students;

[0098] Specifically, a learning behavior vector essentially represents the combination of a student's learning behavior characteristics across multiple dimensions within a time window. Specifically, within each time window, raw feature data across multiple behavioral dimensions is extracted from the student's continuous learning behavior. Based on the nature of each feature (e.g., numerical, proportional, or categorical), corresponding feature standardization or normalization strategies are applied to convert it into a feature vector with a unified dimension. Ultimately, the processed multi-dimensional learning behavior characteristics within that time window constitute a learning behavior vector.

[0099] S2-3, compare the second stage scores with the excellent score threshold and the passing score threshold respectively;

[0100] If the second stage score is higher than the excellent score threshold, the first category student or the second category student to whom the second stage score belongs will be marked as a third category student;

[0101] The third category of students includes the following two categories:

[0102] The first category is: samples with excellent historical grades (i.e., students belonging to the first category), whose grades in the second stage are still above the excellent grade threshold, indicating that their learning behavior and academic performance are continuously consistent;

[0103] The other category is: those with poor history grades (i.e., students belonging to the second category), but whose grades in the second stage have significantly improved to an excellent level, indicating that they have made significant progress in their second stage studies.

[0104] If the second stage score is lower than the passing score threshold, the first category student or the second category student to which the second stage score belongs is marked as a fourth category student;

[0105] The fourth category of students also includes the following two situations:

[0106] The first category is: samples with poor history scores (i.e., students belonging to the second category), and whose scores in the second stage are still below the passing score threshold, indicating that their learning effect continues to be poor;

[0107] The other category is: students with excellent historical grades (i.e. belonging to the first category), but their grades in the second stage dropped significantly to failing, indicating that they may have abnormal fluctuations or regressive behavior in the current stage of study.

[0108] S2-4, from the learning behavior vectors of the first and second types of students, respectively select M learning behavior vectors marked as the third and fourth types of students;

[0109] S2-5, sorting the M learning behavior vectors of the third and fourth types of students respectively to generate corresponding high-performance behavior sequences and low-performance behavior sequences;

[0110] S2-6. Repeat S2-1 to S1-5 until K high-performance behavior sequences and Q low-performance behavior sequences are obtained.

[0111] Furthermore, step S2-2 specifically includes:

[0112] S2-2-1. In the second stage, a fixed-length and sliding time series window is defined; the time series window is used to extract the students' learning behavior data in segments.

[0113] S2-2-2, sliding the time series window with a set sliding step size, and sequentially obtaining the learning behavior vectors of the first and second types of students in the second stage within M time series windows;

[0114] A multi-dimensional behavioral feature vector composed of the response time, page dwell time, click frequency, error rate distribution, and behavioral pause time within each time window.

[0115] S2-2-3. When the time window slides to the end of the second phase, collect the second phase scores of the first and second category students. The second phase scores can be used to correlate learning behavior with learning performance. For example, the second phase scores can be the target student's final scores for this semester.

[0116] Furthermore, step S2-5 specifically includes:

[0117] S2-5-1. Obtain the central timestamps of M time series windows; used to indicate the position of each time series window in the second stage.

[0118] S2-5-2. Encode the central timestamps of the M time windows into time series numbers to clarify the temporal sequence of each learning behavior vector.

[0119] S2-5-3. Sort the learning behavior vectors of the third and fourth categories of students respectively according to the timing numbers of the M timing windows to obtain the high-performance behavior sequence and the low-performance behavior sequence.

[0120] In this embodiment, the set formed by sequentially arranging multiple time windows constitutes a complete behavior sequence. The high-performance behavior sequence and low-performance behavior sequence, respectively, refer to ordered vector sequences formed by concatenating learning behavior vectors within multiple consecutive time windows according to their time sequence numbers. These sequences describe the dynamic characteristic trajectory of the learner's learning behavior over time during the second phase. This behavior sequence not only preserves the multidimensional properties of behavioral characteristics but also reflects temporal dependencies.

[0121] Exemplarily, the steps for screening non-extremely abnormal students include:

[0122] S6-1. Calculate the ranking deviations between the Q reconstructed behavior sequences and the corresponding low-performance behavior sequences to obtain Q ranking deviations for the fourth type of students;

[0123] The expression of the sorting deviation is:

[0124] ;

[0125] in, is the sorting deviation, which is used to measure the difference between the reconstructed behavior sequence and the low-performance behavior sequence in the sorting structure of the behavior vector learned within the time window; M is the number of time windows, is the ranking rank of the low-performance behavior sequence in the t-th time window, Reconstruct the sorting rank in the behavior sequence within the t-th time window;

[0126] Specifically, the sorting rank is obtained as follows:

[0127] First, the learning behavior vector in each time window Calculate their Euclidean norms respectively:

[0128] ;

[0129] in, is the learning behavior vector in the low-performance behavior sequence in the t-th time window, D is the number of feature dimensions of the learning behavior vector, is the eigenvalue of the dth dimension in the tth time window;

[0130] Then for all the Euclidean norms in the sequence Sort and get its sorting number (1-M) in the sequence, which is the sorting rank ; The same method is used to calculate the reconstructed behavior sequence .

[0131] Therefore, although the entire sequence is a whole, the sorting operation is performed based on the vector modulus in each time window, so the sorting rank reflects the intensity or relative position of the behavioral features of each time step.

[0132] S6-2. From the Q ranking deviations, select J ranking deviations that are lower than the ranking deviation threshold;

[0133] S6-3. Mark the fourth category of students to which J ranking deviations belong as non-extreme abnormal students;

[0134] Specifically, non-extremely abnormal students are those whose second-stage scores fall short of passing, but whose learning behavior exhibits relatively stable and orderly temporal evolution. While these students' learning behavior patterns differ somewhat from those of high-performing students, their behavioral sequences exhibit a high degree of temporal consistency and continuity, reflecting the likelihood of possessing a degree of motivation or goal orientation during the learning process. Their current performance falls short of expectations solely due to learning path selection, strategic deviations, or periodic adaptation issues.

[0135] Exemplarily, the steps for constructing a self-supervised training set include:

[0136] S7-1. Filter out J reconstructed behavior sequences of students marked as non-extremely abnormal;

[0137] S7-2. Calculate the reconstruction deviations between the J reconstructed behavior sequences and the corresponding low-performance behavior sequences to obtain J reconstruction deviations of non-extremely abnormal students; the reconstruction deviations are used to measure the degree of deviation between the behavior trajectory of non-extremely abnormal students and the high-performance behavior model.

[0138] The expression of the reconstruction deviation is:

[0139] ;

[0140] in, Reconstruction bias is used to measure the degree of difference in behavior patterns between low-performance behavior sequences and reconstructed behavior sequences. is the learning behavior vector in the low-performance behavior sequence in the t-th time window, is the reconstructed behavior vector in the reconstructed behavior sequence within the t-th time window; is the square of the Euclidean distance between the learned behavior vector in the low-performance behavior sequence and the reconstructed behavior vector in the reconstructed behavior sequence in all feature dimensions, indicating the behavioral deviation in the corresponding time window;

[0141] S7-3. Sort the J reconstruction deviations by size, and assign J abnormality level labels to the low-performance behavior sequences of the non-extremely abnormal students in sequence; the abnormality level label of the smallest reconstruction deviation is set to 1;

[0142] That is, the abnormality level label is numerically positively correlated with the reconstruction deviation, and is used to reflect the degree of deviation between the behavior sequence and the high-performance behavior model. Specifically, multiple abnormality level division intervals are preset, and labels of corresponding levels are assigned according to the interval position of the reconstruction deviation corresponding to each low-performance behavior sequence; for example, multiple discrete level labels can be divided, such as mild abnormality (1), moderate abnormality (2), and severe abnormality (3).

[0143] S7-4, uniformly setting the abnormality level labels of the K high-performance behavior sequences of the third-category students to 0, to form corresponding K abnormality level labels;

[0144] Specifically, the high-performance behavior sequences of the K third-category students have been used to train the temporal behavior reconstruction model in the previous steps. Their high-performance behavior sequences can be regarded as ideal behaviors. Therefore, in the abnormal level recognition task, their corresponding labels are directly set to 0, indicating that there is no deviation in this type of behavior.

[0145] S7-5. Pair J abnormality level labels with corresponding low-performance behavior sequences, and pair K abnormality level labels with corresponding high-performance behavior sequences to construct a self-supervised training set containing S abnormal supervision samples; where S = J + K;

[0146] The anomaly level labeling system constructed in this way is composed of normal samples (label level 0) and anomaly samples with different degrees of deviation (label levels 1, 2, and 3), forming a supervised classification labeling structure with hierarchical distinction capabilities.

[0147] Specifically, each sample in the self-supervised training set consists of a "behavior sequence" and a "corresponding abnormality level label". The behavior sequence is a standardized vector sequence after splicing the time window, and the abnormality level label has a value range of {0, 1, 2, 3}, representing normal, mild, moderate and severe abnormality levels, respectively.

[0148] This embodiment demonstrates a method for identifying and prompting teaching anomalies in teaching robots, overcoming two major technical bottlenecks in existing teaching behavior recognition methods: First, traditional methods rely heavily on manual labeling of abnormal behavior labels, making them difficult to adapt to rapid deployment and dynamic updates in large-scale teaching scenarios; second, existing recognition models often only fit behavior from a single dimension, making it difficult to simultaneously capture deviations in learning behavior across both temporal structure and feature space. By introducing self-supervised label generation and a dual-metric fitting method (ranking bias and reconstruction bias), the accuracy of anomaly recognition is significantly improved.

[0149] The embodiments of the present invention further provide a teaching abnormality identification prompt system for a teaching robot, which is used to implement the above-mentioned method embodiments, and the details already described will not be repeated here. The terms "module," "unit," "subunit," etc. used below may refer to a combination of software and / or hardware that implements a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0150] like Figure 4 As shown, Figure 4 This is a structural block diagram of a teaching abnormality recognition prompt system for a teaching robot according to the present invention, the system comprising:

[0151] Vector acquisition module, used to obtain the real-time learning behavior vector of the student to be identified within the set time window

[0152] An abnormality level output module, used to input the real-time learning behavior vector into a pre-trained abnormality level recognition model and output an abnormality level label corresponding to the real-time learning behavior vector;

[0153] The training of the abnormal level recognition model is based on the sorting deviation and reconstruction deviation of the historical behavior sequence to fit the evolution of the learning behavior vector in the historical data.

[0154] The prompt module is used to trigger the teaching robot's teaching exception prompt according to the exception level label.

[0155] In the above system, the real-time learning behavior vector is obtained through the vector acquisition module; the abnormality level label is obtained through the abnormality level output module; and the teaching abnormality prompt of the teaching robot is triggered through the prompt module, thereby solving the problem of ignoring sorting deviation.

[0156] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiments of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired method (e.g., infrared, wireless, microwave, etc.).

[0157] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A teaching abnormality recognition and prompting method for a teaching robot, characterized in that: include: Obtain the real-time learning behavior vector of the student to be identified within the set time window; Inputting the real-time learning behavior vector into a pre-trained abnormality level recognition model, and outputting an abnormality level label corresponding to the real-time learning behavior vector; The training of the anomaly level recognition model is based on the sorting deviation and reconstruction deviation of the historical behavior sequence to fit the evolution of the learning behavior vector in the historical data; According to the abnormality level label, the teaching robot triggers the teaching abnormality prompt; The modeling steps of the abnormal level identification model include: S1. Obtain the first-stage scores of N target students; S2. Construct a historical behavior sequence based on the first-stage scores of N target students; the historical behavior sequence includes: K high-performance behavior sequences and Q low-performance behavior sequences; S3. defining the K high-performance behavior sequences as an autoencoding training set; S4. Iteratively train the temporal autoencoder model using the autoencoder training set to obtain a temporal behavior reconstruction model that can fit high-performance behavior sequences; S5. Input Q low-performance behavior sequences into the temporal behavior reconstruction model and output Q reconstructed behavior sequences; S6. Based on the Q reconstructed behavior sequences and the corresponding low-performance behavior sequences, select non-extremely abnormal students from the target students; S7. Use high-performance behavior sequences and low-performance behavior sequences of non-extremely abnormal students to construct a self-supervised training set; S8. Use the self-supervised training set to perform supervised training on the time series modeling model to obtain the anomaly level recognition model.

2. The teaching abnormality recognition and prompting method of a teaching robot according to claim 1 is characterized in that: The steps for constructing K high-performance behavior sequences and Q low-performance behavior sequences include: S2-1. Compare the first-stage scores of the target students with the excellent score threshold and the passing score threshold respectively; If the target student's first-stage score is higher than the excellent score threshold, the target student will be marked as a first-category student; If the target student's first-stage score is lower than the passing score threshold, the target student will be marked as a second-category student; S2-2, obtaining the second-stage scores and the second-stage learning behavior vectors of the first and second-stage students; S2-3. Compare the second stage scores with the excellent score threshold and the passing score threshold respectively; If the second stage score is higher than the excellent score threshold, the first category student or the second category student to whom the second stage score belongs will be marked as a third category student; If the second stage score is lower than the passing score threshold, the first category student or the second category student to which the second stage score belongs is marked as a fourth category student; S2-4, from the learning behavior vectors of the first and second types of students, respectively select M learning behavior vectors marked as the third and fourth types of students; S2-5, sorting the M learning behavior vectors of the third and fourth types of students respectively to generate corresponding high-performance behavior sequences and low-performance behavior sequences; S2-6. Repeat S2-1 to S2-5 until K high-performance behavior sequences and Q low-performance behavior sequences are obtained.

3. The teaching abnormality recognition and prompting method of a teaching robot according to claim 2, characterized in that: Obtaining the second-stage scores and the second-stage learning behavior vectors of the first and second-stage students includes: S2-2-1. In the second stage, a fixed-length and sliding timing window is defined; S2-2-2, sliding the time series window with a set sliding step size, and sequentially obtaining the learning behavior vectors of the first and second types of students in the second stage within M time series windows; S2-2-3. When the timing window slides to the end of the second stage, the second stage scores of the first and second types of students are collected.

4. The teaching abnormality recognition and prompting method of a teaching robot according to claim 3 is characterized in that: Generating corresponding high-performance behavior sequences and low-performance behavior sequences includes: S2-5-1. Obtain the central timestamps of M time series windows; S2-5-2. Encode the central timestamps of the M time series windows into time series numbers; S2-5-3. Sort the learning behavior vectors of the third and fourth categories of students respectively according to the timing numbers of the M timing windows to obtain the high-performance behavior sequence and the low-performance behavior sequence.

5. The teaching abnormality recognition and prompting method of a teaching robot according to claim 4, characterized in that: Screening non-extreme abnormal students from target students includes: S6-1. Calculate the ranking deviations between the Q reconstructed behavior sequences and the corresponding low-performance behavior sequences to obtain Q ranking deviations for the fourth type of students; The expression of the sorting deviation is: ； in, is the sorting deviation, M is the number of time windows, is the ranking rank of the low-performance behavior sequence in the t-th time window, Reconstruct the sorting rank in the behavior sequence within the t-th time window; S6-2. From the Q ranking deviations, select J ranking deviations that are lower than the ranking deviation threshold; S6-3. Mark the fourth category of students to which J ranking deviations belong as non-extreme abnormal students.

6. The teaching abnormality recognition and prompting method of a teaching robot according to claim 5, characterized in that: Constructing a self-supervised training set includes: S7-1. Filter out J reconstructed behavior sequences of students marked as non-extremely abnormal; S7-2. Calculate the reconstruction deviations between the J reconstructed behavior sequences and the corresponding low-performance behavior sequences to obtain J reconstruction deviations for non-extremely abnormal students; The expression of the reconstruction deviation is: ； in, To reconstruct the deviation, is the learning behavior vector in the low-performance behavior sequence in the t-th time window, is the reconstructed behavior vector in the reconstructed behavior sequence within the t-th time window; is the square of the Euclidean distance between the learned behavior vector in the low-performance behavior sequence and the reconstructed behavior vector in the reconstructed behavior sequence in all feature dimensions; S7-3. Sort the J reconstruction deviations by size, and assign J abnormality level labels to the low-performance behavior sequences of the non-extremely abnormal students in sequence; the abnormality level label of the smallest reconstruction deviation is set to 1; S7-4, uniformly setting the abnormality level labels of the K high-performance behavior sequences of the third-category students to 0, to form corresponding K abnormality level labels; S7-5. Pair J anomaly level labels with corresponding low-performance behavior sequences, and pair K anomaly level labels with corresponding high-performance behavior sequences to construct a self-supervised training set containing S anomaly supervision samples; where S = J + K.

7. A teaching abnormality recognition and prompting system for a teaching robot, comprising: Vector acquisition module, used to obtain the real-time learning behavior vector of the student to be identified within the set time window An abnormality level output module, used to input the real-time learning behavior vector into a pre-trained abnormality level recognition model and output an abnormality level label corresponding to the real-time learning behavior vector; The training of the abnormal level recognition model is based on the sorting deviation and reconstruction deviation of the historical behavior sequence to fit the evolution of the learning behavior vector in the historical data. The prompt module is used to trigger the teaching robot's teaching abnormality prompt according to the abnormality level label; The modeling steps of the abnormal level identification model include: S1. Obtain the first-stage scores of N target students; S2. Construct a historical behavior sequence based on the first-stage scores of N target students; the historical behavior sequence includes: K high-performance behavior sequences and Q low-performance behavior sequences; S3. defining the K high-performance behavior sequences as an autoencoding training set; S4. Iteratively train the temporal autoencoder model using the autoencoder training set to obtain a temporal behavior reconstruction model that can fit high-performance behavior sequences; S5. Input Q low-performance behavior sequences into the temporal behavior reconstruction model and output Q reconstructed behavior sequences; S6. Based on the Q reconstructed behavior sequences and the corresponding low-performance behavior sequences, select non-extremely abnormal students from the target students; S7. Use high-performance behavior sequences and low-performance behavior sequences of non-extremely abnormal students to construct a self-supervised training set; S8. Use the self-supervised training set to perform supervised training on the time series modeling model to obtain the anomaly level recognition model.

Images