A scoring method, apparatus, device, and computer-readable storage medium

Abstract

This application discloses a scoring method, apparatus, device, and computer-readable storage medium. The method includes: acquiring identification information of an assessment subject and voice evaluation information for the assessment subject, wherein the voice evaluation information includes at least behavioral feature information of the assessment subject; acquiring a trained scoring model; using the trained scoring model to score the voice evaluation information to obtain the assessment target of the assessment subject and the scoring result corresponding to the assessment target; and outputting the identification information of the assessment subject, the assessment target, and the scoring result corresponding to the assessment target. Automatic scoring can be achieved based on the trained scoring model without the need for manual input of scores, thereby improving the flexibility and automation of scoring and thus increasing scoring efficiency.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and includes, but is not limited to, a scoring method, apparatus, device, and computer-readable storage medium. Background Technology

[0002] With the continuous promotion of paperless office, the moral education evaluation management system has emerged and has become one of the more commonly used digital assessment systems in the primary and secondary education stage. It enables school teachers and administrators to quickly evaluate students' moral education through the system, thereby improving management efficiency.

[0003] The operational process of the relevant moral education evaluation management system is as follows: First, the administrator configures assessment indicators in the backend, such as assessment items and scoring standards for hygiene, discipline, eye exercises, etiquette, and safety. Then, in actual use, teachers select the list of students to be assessed on the frontend and manually fill in the corresponding evaluation content and scores for the assessment items according to the assessment indicators. The frontend submits the evaluation content and scores for the assessment items to the backend. Finally, the backend summarizes and analyzes the scores, calculates the total scores and rankings for different dimensions, and pushes the total scores and rankings for different dimensions to the administrator or teacher.

[0004] Although relevant moral education evaluation management systems have achieved paperless storage and statistics through digital means, in actual operation, they cannot evaluate students in a timely manner and still require manual input of information to complete the evaluation. This undoubtedly takes up a lot of teachers' working time, affects normal teaching activities, makes the evaluation untimely, and has low flexibility and automation, thus reducing evaluation efficiency. Summary of the Invention

[0005] In view of the above, embodiments of this application provide a scoring method, apparatus, device, and computer-readable storage medium.

[0006] The technical solution of this application embodiment is implemented as follows:

[0007] This application provides a scoring method, including:

[0008] Obtain the identification information of the assessment subject and the voice evaluation information of the assessment subject, wherein the voice evaluation information includes at least the behavioral characteristic information of the assessment subject;

[0009] Obtain the trained scoring model;

[0010] The trained scoring model is used to score the speech evaluation information to obtain the assessment target of the assessment object and the corresponding score result of the assessment target.

[0011] Output the identification information of the assessment object, the assessment target, and the scoring result corresponding to the assessment target.

[0012] This application provides a scoring device, the device comprising:

[0013] The first acquisition module is used to acquire the identification information of the assessment object and the voice evaluation information of the assessment object, wherein the voice evaluation information includes at least the behavioral characteristic information of the assessment object.

[0014] The second acquisition module is used to acquire the trained scoring model;

[0015] The scoring module is used to process the speech evaluation information using the trained scoring model to obtain the assessment target of the assessment object and the scoring result corresponding to the assessment target.

[0016] The output module is used to output the identification information of the assessment object, the assessment target, and the scoring result corresponding to the assessment target.

[0017] This application embodiment provides a scoring device, the scoring device comprising:

[0018] Processor; and

[0019] Memory for storing computer programs that can run on the processor;

[0020] The computer program implements the scoring method described above when executed by a processor.

[0021] This application provides a computer-readable storage medium storing computer-executable instructions configured to execute the above-described scoring method.

[0022] This application provides a scoring method, apparatus, device, and computer-readable storage medium. The method acquires the identification information of the assessment subject and voice evaluation information for the assessment subject, wherein the voice evaluation information includes at least the behavioral characteristic information of the assessment subject. Next, a trained scoring model is acquired, and the voice evaluation information is processed using the trained scoring model to obtain the assessment target of the assessment subject and the corresponding scoring result. Finally, the identification information of the assessment subject, the assessment target, and the corresponding scoring result are output. This allows for timely scoring of the assessor based on the assessment subject's identification information, voice evaluation information, and the trained scoring model, without being limited by geographical location, thus improving the real-time nature of the scoring. Furthermore, automatic scoring can be achieved based on the trained scoring model without manual input of scores, thereby improving the flexibility and automation of the scoring process and ultimately increasing scoring efficiency. Attached Figure Description

[0023] In the accompanying drawings (which are not necessarily drawn to scale), similar reference numerals may describe similar parts in different views. The drawings illustrate, by way of example and not limitation, the various embodiments discussed herein.

[0024] Figure 1 This is a schematic diagram illustrating an implementation process of the scoring method provided in an embodiment of this application;

[0025] Figure 2 A schematic diagram illustrating an implementation process of a method for continuing to train a pre-trained scoring model, as provided in an embodiment of this application.

[0026] Figure 3 A schematic diagram illustrating another implementation flow of the scoring method provided in the embodiments of this application;

[0027] Figure 4 A schematic diagram illustrating an implementation process of a method for obtaining the assessment target of an assessment object through a trained scoring model, as provided in an embodiment of this application.

[0028] Figure 5 A schematic diagram illustrating another implementation of the scoring method provided in the embodiments of this application;

[0029] Figure 6 This is a schematic diagram of the composition structure of the moral education evaluation and management system provided in the embodiments of this application;

[0030] Figure 7 This is a schematic diagram of the composition of the scoring device provided in the embodiments of this application;

[0031] Figure 8 This is a schematic diagram of the composition structure of the scoring device provided in the embodiments of this application. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0034] In the following description, the terms "first, second, third" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, third" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0036] To address the problems existing in related technologies, this application provides a scoring method. This method can be implemented by a computer program, which, when executed, completes each step of the scoring method. In some embodiments, the computer program can be executed on a processor in a scoring device. Figure 1 This is a schematic diagram illustrating an implementation flow of the scoring method provided in an embodiment of this application, such as... Figure 1 As shown, the method includes:

[0037] Step S101: Obtain the identification information of the assessment subject and the voice evaluation information for the assessment subject.

[0038] Here, the voice evaluation information includes at least the behavioral characteristic information of the assessment subject. The assessment subject can be a school student, company employee, taxi driver, or other individuals requiring assessment. Furthermore, the behavioral characteristic information is used to characterize a specific behavior of the assessment subject or a characteristic of the subject themselves. For example, if the assessment subject is a school student, the behavioral characteristic information could include sleeping in class, not wearing a school uniform, or voluntarily picking up litter. In addition, voice evaluation information refers to information presented in voice form; for example, the evaluator might directly say "sleeping in class."

[0039] In this embodiment, step S101 can be performed by a general-purpose server or a dedicated server. Furthermore, the voice evaluation information can originate from terminal devices such as smartphones or smart computers. To reduce the transmission bit rate, save storage space, and improve transmission or storage efficiency, the terminal device encodes the voice evaluation information and sends the encoded voice evaluation information to the server. That is, the server receives the encoded voice evaluation information. The encoding of the voice evaluation information can employ waveform encoding, parametric encoding, or a hybrid encoding method.

[0040] In this embodiment of the application, the identification information of the assessment object can be the name, number, etc. of the assessment object. Taking a school student as an example, the identification information can be the student's name, or the student's student ID, etc.

[0041] Step S102: Obtain the trained scoring model.

[0042] Here, the trained scoring model can be an artificial intelligence algorithm model such as a trained neural network model, a trained Bayesian network model, or a trained genetic algorithm model. This trained scoring model can determine the assessment target of the assessment object and the corresponding scoring result based on the voice evaluation information, thereby achieving the purpose of automatic and intelligent scoring.

[0043] In some embodiments, a preset scoring model needs to be trained before step S102 to obtain a trained scoring model. In implementation, a preset scoring model can be obtained first, which may be an artificial intelligence algorithm model such as a neural network model, a Bayesian network model, or a genetic algorithm model. Then, sample speech evaluation information is input into the preset scoring model to obtain the predicted assessment target and the predicted score result corresponding to the sample speech evaluation information. Next, it is determined whether the sample assessment target and the predicted assessment target meet a preset matching condition. Here, the matching condition can be that the sample assessment target and the predicted assessment target are consistent. When it is determined that the sample assessment target and the predicted assessment target meet the preset matching condition, error information between the sample score result and the predicted score result is further obtained. The error information can be the absolute value of the difference between the sample score result and the predicted score result. Finally, the preset scoring model is backpropagated and trained based on the error information and a preset error threshold to obtain a trained scoring model.

[0044] Step S103: Use the trained scoring model to score the speech evaluation information to obtain the assessment target of the assessment object and the scoring result corresponding to the assessment target.

[0045] In this embodiment of the application, the text data corresponding to the voice scoring model can be obtained first, and the text data can be input into the trained scoring model. The trained scoring model can then output the assessment target of the assessment object and the scoring result corresponding to the assessment target, thereby obtaining the assessment target of the assessment object and the scoring result corresponding to the assessment target. For example, the assessment target of the assessment object can be discipline, and the scoring result corresponding to the assessment target can be 8 points.

[0046] In some embodiments, a set of reference assessment targets is determined based on the identification information of the assessment target. For example, if the identification information of the assessment target is "first grade", then the set of assessment targets corresponding to "first grade" is determined as the set of reference assessment targets. If the set of assessment targets corresponding to "first grade" is hygiene, discipline, and etiquette, then hygiene, discipline, and etiquette are determined as the set of reference assessment targets. Further, it is determined whether the set of reference assessment targets includes assessment targets. If the set of reference assessment targets includes assessment targets, it indicates that the obtained assessment targets are correct; if the set of reference assessment targets does not include assessment targets, it indicates that the obtained assessment targets are incorrect. In this case, it is necessary to determine the reference assessment items associated with the speech evaluation information from the set of reference assessment targets. Finally, the trained scoring model is further trained based on the speech evaluation information and the reference assessment targets to obtain a retrained scoring model. This continuously improves the trained scoring model, so that when speech evaluation information is input into the trained scoring model again, the retrained scoring model can output the correct assessment targets, thereby correctly associating assessment targets with more speech evaluation information. When determining reference assessment targets, a similarity algorithm can be used to identify the reference assessment target with the highest similarity to the speech evaluation information, or the reference assessment targets included in the reference assessment target set can be directly received to obtain the reference assessment targets.

[0047] Step S104: Output the identification information of the assessment object, the assessment target, and the scoring result corresponding to the assessment target.

[0048] In this embodiment of the application, the identification information of the assessment object, the assessment target, and the scoring result corresponding to the assessment target can be returned to the terminal device so that the identification information of the assessment object, the assessment target, and the scoring result corresponding to the assessment target can be displayed on the terminal.

[0049] This application provides a scoring method that acquires the identification information of the assessment subject and the voice evaluation information for the assessment subject, wherein the voice evaluation information includes at least the behavioral characteristic information of the assessment subject; then, a trained scoring model is acquired, and the trained scoring model is used to process the voice evaluation information to obtain the assessment target of the assessment subject and the corresponding scoring result; finally, the identification information of the assessment subject, the assessment target, and the corresponding scoring result are output. This allows for timely scoring of the assessor based on the identification information, voice evaluation information, and the trained scoring model, without being limited by geographical location, thus improving the real-time performance of the scoring; furthermore, automatic scoring can be achieved based on the trained scoring model without requiring manual input of scores, thereby improving the flexibility and automation of the scoring process, and thus increasing scoring efficiency.

[0050] Based on the above embodiments, this application further provides a scoring method, such as... Figure 2 As shown, after step S103, the method further includes the following steps S201 to S205:

[0051] Step S201: Based on the identification information of the assessment object, determine the set of reference assessment targets corresponding to the assessment object.

[0052] Here, the server stores the pre-configured mapping relationship between the assessment object identification information and the corresponding reference assessment target set. For example, if the assessment object identification information is first grade, and the corresponding assessment target set for first grade is hygiene, discipline, and etiquette, then the reference assessment target set is hygiene, discipline, and etiquette. If the assessment object identification information is seventh grade, and the corresponding reference assessment target set for seventh grade is discipline, social practice, and class attendance, then the reference assessment target set is discipline, social practice, and class attendance.

[0053] Step S202: Determine whether the reference assessment target set includes assessment targets.

[0054] Here, when the reference assessment target set includes the assessment target, it indicates that the trained scoring model can evaluate the correct assessment target. In this case, proceed to step S203, that is, maintain the trained scoring model. When the reference assessment target set does not include the assessment target, it indicates that the trained scoring model cannot evaluate the correct assessment target, and it is necessary to continue training the trained scoring model. In this case, proceed to step S204.

[0055] Step S203: Keep the trained scoring model unchanged.

[0056] Here, there is no need to retrain the already trained scoring model; the trained scoring model can be maintained.

[0057] Step S204: Determine the reference assessment targets associated with the voice evaluation information from the set of reference assessment targets.

[0058] Here, similarity algorithms such as Manhattan distance, Euclidean distance, and Chebyshev distance can be used to calculate the similarity between each reference assessment target in the reference assessment target set and the speech evaluation information. The reference assessment target with the highest similarity is then determined as the reference assessment target associated with the speech evaluation information.

[0059] In some embodiments, a reference assessment target associated with voice evaluation information can be determined in the following ways: information for inputting the reference assessment target is sent to the terminal device, and the reference assessment target is directly received from the terminal device.

[0060] Step S205: Based on the voice evaluation information and the reference assessment target, continue to train the trained scoring model to obtain a retrained scoring model.

[0061] Here, the trained scoring model can be further trained using gradient descent or adaptive moment estimation to obtain a retrained scoring model. This allows the trained scoring model to be continuously improved. When the speech evaluation information is input into the trained scoring model again, the retrained scoring model can output the correct assessment target, thereby correctly associating more speech evaluation information with the assessment target.

[0062] Through steps S201 to S205 above, a set of reference assessment targets corresponding to the assessment target is determined based on the identification information of the assessment target. Then, it is determined whether the set of reference assessment targets includes assessment targets. If the set of reference assessment targets includes assessment targets, it indicates that the trained scoring model can evaluate the correct assessment targets, and the trained scoring model is maintained. If the set of reference assessment targets does not include assessment targets, it indicates that the trained scoring model cannot evaluate the correct assessment targets, and the trained scoring model needs to be trained again. Next, reference assessment targets associated with speech evaluation information are determined from the set of reference assessment targets. Finally, the trained scoring model is trained again based on speech evaluation information and reference assessment targets to obtain a retrained scoring model, so that the trained scoring model is continuously improved, thereby enabling it to correctly associate assessment targets with more speech evaluation information, thus increasing the robustness of the trained scoring model.

[0063] Based on the above embodiments, this application further provides a scoring method, applied to terminal devices and servers, such as... Figure 3 As shown, the method includes the following steps S301 to S316:

[0064] Step S301: The server obtains the preset scoring model, sample voice evaluation information, sample assessment target corresponding to the sample voice evaluation information, and sample scoring result corresponding to the sample assessment target.

[0065] Here, the preset scoring model can be an artificial intelligence algorithm model such as a neural network model, a Bayesian network model, or a genetic algorithm model. The sample scoring result can be a score for the sample assessment target. There is a correspondence between the sample voice evaluation information, the sample assessment target corresponding to the sample voice evaluation information, and the sample scoring result corresponding to the sample assessment target. For example, if the sample voice evaluation information is "sleeping in class", then the sample assessment target can be "discipline", and the sample scoring result is 5 points.

[0066] In this embodiment of the application, a terminal device can receive a preset scoring model, sample voice evaluation information, sample assessment target corresponding to the sample voice evaluation information, and sample scoring result corresponding to the sample assessment target. It can also directly receive the preset scoring model, sample voice evaluation information, sample assessment target corresponding to the sample voice evaluation information, and sample scoring result corresponding to the sample assessment target, and store them in its own storage space.

[0067] In step S302, the server uses the preset scoring model to score the sample voice evaluation information, and obtains the predicted assessment target and the predicted score result of the predicted assessment target corresponding to the sample voice evaluation information.

[0068] In this embodiment, the sample voice evaluation information can be decoded, sampled, and recognized to obtain sample text data. Then, the sample text data can be input into a preset scoring model. The preset scoring model can output the predicted assessment target and the predicted score result corresponding to the sample voice evaluation information. For example, if the sample voice evaluation information is "sleeping in class", the preset scoring model can obtain the predicted sample assessment target as "discipline" and the predicted score result as 6 points.

[0069] In step S303, when the server determines that the sample assessment target and the predicted assessment target meet the preset matching conditions, it obtains the error information between the sample scoring result and the predicted scoring result.

[0070] Here, the matching condition can be a consistency condition, meaning it will determine whether the sample assessment target is consistent with the predicted assessment target. If the sample assessment target is inconsistent with the predicted assessment target, the preset scoring model needs to be further trained using the sample voice evaluation information and the sample assessment target, so that the trained scoring model can make the predicted assessment target consistent with the sample assessment target. If the sample assessment target is consistent with the predicted assessment target, as in the example above where both the sample assessment target and the predicted assessment target are "discipline," then the error information between the sample scoring result and the predicted scoring result is further obtained. When obtaining the error information, it can be the absolute value of the difference between the sample scoring result and the predicted scoring result. Continuing with the example above, if the sample scoring result is 5 points and the predicted scoring result is 6 points, then the error information is 1 point.

[0071] In step S304, the server performs backpropagation training on the preset scoring model based on the error information and the preset error threshold to obtain the trained scoring model.

[0072] Here, the preset error threshold can be 0.5 points, 1 point, 1.5 points, etc. Taking an error information of 1 point and a preset error threshold of 0.5 points as an example, if the error information is greater than the error threshold, the error information greater than the error threshold will be backpropagated to determine the gradient vector. Then, each weight in the preset scoring model will be adjusted through the gradient vector to train the preset scoring model until the error information is less than or equal to the preset error threshold, thus obtaining the trained scoring model.

[0073] In step S305, the terminal device sends the image of the assessment subject to be recognized and the voice evaluation information for the assessment subject to the server.

[0074] Here, the terminal device can obtain the image of the subject to be identified in a timely manner through its own image acquisition device, and can also obtain the voice evaluation information of the subject to be identified in a timely manner through its own audio acquisition device. Furthermore, in order to reduce the transmission bit rate, save storage space, and improve transmission or storage efficiency, the terminal device will encode the voice evaluation information; finally, the image to be identified and the encoded voice evaluation information will be sent to the server.

[0075] In step S306, the server determines the facial image from the image to be recognized and extracts the facial feature information of the facial image.

[0076] Here, the image to be identified includes at least a facial image. In practice, the image to be identified may also include background images such as the ground and trees, as well as images of the neck and upper body of the subject. Therefore, it is necessary to determine the facial image from the image to be identified using methods such as gradient vector flow and deformed perpendicular curves. Then, facial image feature information can be extracted from the facial image using methods such as histogram of oriented gradients and convolutional neural networks.

[0077] In step S307, the server determines whether the identification information of the assessment subject can be determined from the preset information database based on the facial feature information.

[0078] Here, the preset information database stores the correspondence between facial feature information and identification information. Next, the facial feature information of the assessment subject is compared with the facial feature information in the preset information database to determine whether the facial feature information in the preset information database includes the facial feature information of the assessment subject. If the facial feature information in the preset information database includes the facial feature information of the assessment subject, it is considered that the identification information of the assessment subject can be determined from the preset information database, and the voice evaluation information is further processed, proceeding to step S308; if the facial feature information in the preset information database does not include the facial feature information of the assessment subject, it is considered that the identification information of the assessment subject cannot be determined from the preset information database, and the facial feature information and identification information of the assessment subject need to be added to the preset information database, proceeding to step S312.

[0079] In step S308, the server decodes the voice evaluation information to obtain the decoded voice evaluation information.

[0080] Here, after obtaining the voice evaluation information, the server will use a decoding method corresponding to the encoding method to decode the voice evaluation information, thereby obtaining the decoded voice evaluation information.

[0081] Step S309: The server samples the decoded speech evaluation information to obtain sampled speech evaluation information.

[0082] Here, sampling processing can remove silence and noise from speech and enhance the effective parts of speech. Endpoint detection technology can be used when performing sampling processing.

[0083] In step S310, the server performs speech recognition on the sampled speech evaluation information using a preset acoustic model to obtain text data.

[0084] Here, the preset acoustic model can be a Gaussian Mixture Model-Hidden Markov Model (GMM-HMM) model, or a Deep Neural Networks-Hidden Markov Model (DNN-HMM) model. The text data corresponding to the speech evaluation information can be obtained through the above preset acoustic models.

[0085] In step S311, the server inputs the text data into the trained scoring model to obtain the assessment target of the assessment object and the corresponding scoring result.

[0086] Here, taking a trained scoring model as an example, we can first perform feature extraction preprocessing on the text data, and then propagate the preprocessed text data forward through the neural network to obtain the scoring result.

[0087] like Figure 4 As shown, inputting text data into the trained scoring model to obtain the assessment objectives for the assessed individuals can be achieved through the following steps S3111 to S3113:

[0088] Step S3111: Input the text data into the trained classification model, and determine the set of similar assessment targets for the text data based on the text data and the first preset search algorithm.

[0089] Here, the first preset search algorithm can be a content-based recommendation algorithm, a recommendation algorithm based on association rules, or a knowledge-based recommendation algorithm, etc. In order to improve the overall efficiency of determining the assessment target, the set of similar assessment targets corresponding to the text data can be determined first through the first preset search algorithm.

[0090] Step S3112: Determine the assessment targets of the text data based on the set of similar assessment targets and the second preset search algorithm.

[0091] Here, the second preset search algorithm can be an artificial intelligence algorithm such as a genetic algorithm or a neural network algorithm, which can quickly determine the assessment target corresponding to the text data from a set of similar assessment targets.

[0092] Step S3113: Determine the assessment objectives of the text data as the assessment objectives of the assessment object.

[0093] Here, the text data describes the behavioral characteristics of the assessment target. Therefore, the assessment target corresponding to the text data is the assessment target of the assessment target.

[0094] Through the above steps S3111 to S3113, a set of similar assessment targets for the text data can be determined first, and then the assessment targets for the text data can be quickly determined from the set of similar assessment targets. Finally, the assessment targets for the text data can be determined as the assessment targets for the assessment objects, thereby improving the speed and accuracy of determining assessment targets.

[0095] In step S312, the server sends a prompt message to the terminal device.

[0096] At this point, if the identification information of the assessment subject cannot be determined from the preset information database, and the facial feature information and identification information of the assessment subject need to be added to the preset information database, the server will output a prompt message and send the prompt message to the terminal device. The prompt message is used to prompt the input of the identification information of the assessment subject.

[0097] Step S313: The terminal device obtains the identification information of the assessment object.

[0098] Here, the terminal device can obtain the identification information of the assessment subject through external devices such as a keyboard and microphone.

[0099] Step S314: The terminal device sends the identification information of the assessment object to the server.

[0100] Here, the terminal device can send the identification information of the assessment object to the server via wired or wireless means.

[0101] Step S315: The server establishes a correspondence between the facial feature information of the assessment subject and the identification information of the assessment subject.

[0102] Here, in order to determine the identification information of the assessment subject based on the subject's facial features in the next assessment, the server will establish a correspondence between the subject's facial features and the subject's identification information after receiving the subject's identification information.

[0103] In step S316, the server stores the corresponding relationship in a preset information database.

[0104] Here, the correspondence between the facial features of the assessment subject and the identification information of the assessment subject is written into the preset information database, thereby enriching the preset information database.

[0105] Through steps S301 to S316, the server decodes, samples, and recognizes the acquired sample speech evaluation information to obtain sample text data; the sample text data is input into a preset scoring model to obtain the predicted assessment target and the predicted score result of the predicted assessment target; further, when the sample assessment target is consistent with the predicted assessment target, the error information between the sample score result and the predicted score result is obtained, and the preset scoring model is backpropagated and trained based on the error information and error threshold to obtain a trained scoring model. Thus, the training of the preset scoring model is completed, and a trained scoring model is obtained; then, the terminal device will... The system recognizes image and voice evaluation information and sends it to the server. On one hand, the server identifies the facial image from the image to be recognized, extracts facial feature information, and then determines the identification information of the assessment subject from a pre-set information database based on the facial feature information. On the other hand, the server decodes, samples, and recognizes the voice evaluation information to obtain text data, which is then input into a trained scoring model. This yields the assessment target for the assessment subject and the corresponding scoring result, enabling automatic scoring based on the trained scoring model without manual input, thus improving the flexibility and automation of the scoring process and increasing efficiency. Furthermore, when the identification information of the assessment subject cannot be determined from the pre-set information database, the system can send a prompt message and receive identification information from the terminal device, establishing a correspondence between the facial feature information of the assessment subject and the identification information of the assessment subject, and storing this information in the pre-set information database, thereby enriching and improving the information in the database.

[0106] Based on the above embodiments, this application provides another scoring method applied to a campus moral education management system. The assessment object is the students being assessed. Although the campus moral education management system achieves paperless storage and statistics of campus moral education management scenarios through digital means, the system still has limitations in realizing responsive functions. These limitations are mainly manifested in: low flexibility in teacher evaluation, inability to evaluate students instantly, and reliance on manual input for student evaluation. This undoubtedly consumes a significant amount of teachers' working time, affecting the normal teaching of the school and its teachers. Therefore, the system's intelligence needs to be improved.

[0107] The moral education scoring method provided in this application applies artificial intelligence (AI) technology to the campus moral education evaluation and management system, enabling teachers to leave real-time evaluation records in any teaching scenario, and providing school teachers with a set of intelligent moral education management tools to help campus administrators and teachers get rid of tedious work, thereby improving management efficiency.

[0108] The moral education scoring method provided in this application embodiment can be implemented through the following process: First, the administrator configures assessment indicators such as hygiene, discipline, eye exercises, etiquette, and safety, as well as assessment items and scoring standards in the backend, and configures language model training and inputs facial information to form a facial information database for evaluation management. Then, the teacher uses a mobile application (APP) client to arbitrarily photograph students' misbehavior, such as spitting or improper exercise, and identifies the current student's information through the facial recognition module. Finally, the teacher inputs evaluation information via voice, such as "a student is sleeping in class." The system can then recognize the language, convert it into system assessment items, such as discipline assessment items, and make an evaluation judgment to obtain the student's discipline score.

[0109] Furthermore, to achieve the above process, the technical solution provided in this application includes the following three parts:

[0110] Part 1: Facial Recognition Database Management

[0111] The following two steps can be performed in the management of the facial information database:

[0112] Step 1: Use a face capture terminal to capture facial images and upload them to the server.

[0113] Step two: The server extracts features from the facial images, stores the feature information in a relational database management system (MySQL) database, and adds identification information for the students being assessed.

[0114] Part Two: Recognition Model Management.

[0115] The following three steps can be performed in the management of recognition models:

[0116] Step 1: Face detection. Determine the size and position of the face in the image and crop the face out of the image.

[0117] Step two, face alignment: Find several key points (reference points, such as the corners of the eyes, the tip of the nose, the corners of the mouth, etc.) of the face from different poses and expressions of the same person, and then use these corresponding key points to transform the face into a standard face as much as possible through similar transformations (rotation, scaling and translation).

[0118] Step 3: Facial feature representation. Vectorized facial features are obtained through feature modeling. Finally, the recognition result is judged by a discriminator. This recognition result is the identification information for assessing students. The recognition algorithm can use Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). PCA is also known as Eigenface.

[0119] Part Three: Speech Model Training.

[0120] The following three steps can be performed during speech model training:

[0121] Step one, voice encoding, uses a hybrid encoding method that combines the principles of waveform encoding and vocoder. The bit rate is approximately between 4kbit / s and 16kbit / s, resulting in better sound quality. This reduces the transmission bit rate or storage requirements, thereby improving transmission or storage efficiency.

[0122] Step two, speech recognition: Speech sampling is performed using endpoint detection technology, and speech recognition is performed using the traditional template matching method. The collected speech data can be input into the GMM-HMM acoustic model and the DNN-HMM acoustic model for speech recognition to obtain the corresponding text data.

[0123] Step three involves voice tagging. Tags are added to the text data and associated with performance indicators. A recommendation algorithm is used to rank and index the similarity between the indicators and text data, recommending the range of indicators associated with the text. Then, a genetic algorithm refines the range of indicators associated with the text down to the specific indicator items. Finally, a deep learning model is trained. Training the deep learning model can be achieved through the following six sub-steps:

[0124] The first sub-step is to preprocess the data.

[0125] In the second sub-step, the processed data is input into the neural network for forward propagation to obtain the score.

[0126] In sub-step three, the score is input into the error function and compared with the expected value to obtain the error. If there are multiple errors, they are summed. The degree of recognition is judged by the error.

[0127] Sub-step four involves determining the gradient vector through backpropagation.

[0128] Sub-step five involves adjusting each weight using the gradient vector to guide the score towards a trend where the error approaches zero or converges.

[0129] Sub-step six: Repeat sub-steps one through five above until the set number of times or the average value of the loss no longer decreases.

[0130] By completing steps one through six above, the training of the deep learning model is finished.

[0131] like Figure 5 As shown, the moral education scoring method provided in this application embodiment can be implemented through the following steps S501 to S508:

[0132] Step S501: Configure moral education evaluation and assessment indicators.

[0133] Step S502: Configure the speech training model.

[0134] Here, the voice data is associated with assessment indicators and then input into the moral education scoring model for training to obtain a trained moral education scoring model. In this embodiment of the application, the trained moral education scoring model can be referred to as the voice training model.

[0135] Step S503: Enter the facial recognition database of all students being assessed.

[0136] Here, the collected facial images are processed by feature extraction and stored in the facial database.

[0137] Step S504: Configure the recognition model parameters.

[0138] Here, open the mobile app, take a picture of your face, and transmit it to the facial recognition server.

[0139] Step S505: Determine whether the client can recognize the identification information of the student being assessed.

[0140] Here, the face recognition server uses the recognition algorithm in the configured resolver to compare the captured face image with the face database to confirm the identification information of the student being assessed. If the student's facial feature information and the student's identification information are found in the face database, the comparison is considered correct, and the process proceeds to S506. If the student's facial feature information and the student's identification information are not found in the face database, the comparison is considered to have failed. The captured face of the student is then re-entered into the face database and marked so that the administrator can manually confirm the identity. After confirmation, the process returns to S503, that is, the face entry is restarted.

[0141] Step S506: Input the voice data for moral education evaluation through the microphone and upload it to the system backend.

[0142] Here, the system backend can be a speech recognition server, which can be the same server as the face recognition server, possessing the functions of both. Further, the speech recognition server decodes the moral education evaluation speech data and converts the decoded speech data into text data using an acoustic model.

[0143] In step S507, the system backend receives the identification information and text data of the student being assessed, and determines whether the text data matches the target assessment item corresponding to the identification information of the student being assessed.

[0144] Here, the transformed text data is input into the trained moral education scoring model. This model matches the text data with the target assessment items corresponding to the students being assessed. If a match is successful, a score is awarded for that assessment indicator. Here, the target assessment item refers to the assessment indicator under different application scenarios. For example, in a grade-level scenario, primary schools have primary school assessment indicators, and secondary schools have secondary school assessment indicators. The process proceeds to step S508, which involves sending the scoring results and statistical results to the client. If a match fails, the moral education evaluation voice data is added to the failed match database. Then, the moral education evaluation voice data in the failed match database is extracted, associated with the assessment indicator items, and the voice model is retrained, returning to step S502.

[0145] Step S508: Return the evaluation score and evaluation score statistics to the mobile APP client.

[0146] Here, the statistical result of the evaluation score can be the sum of the evaluation scores of all students in each class, or the sum of the evaluation scores of all students in each grade.

[0147] Steps S501 to S508 above utilize AI algorithms. On one hand, they accurately locate students being assessed in real time and quickly output scoring results using the advantages of language evaluation, providing teachers with a more intelligent means for moral education evaluation management. On the other hand, they enrich the application of AI technology in the field of smart education. The integration of AI technology enables rapid identification of students being assessed, avoiding the tediousness of traditional paper records or manual input, improving efficiency, and providing administrators with management tools. It allows teachers to quickly and instantly record assessments of the assessed targets in any scenario, and also enables the model setting of moral education management evaluation indicators, applying deep learning training technology to train the moral education scoring model using voice evaluation.

[0148] In some embodiments, such as Figure 6 As shown, the moral education evaluation management system 60 provided in this application embodiment comprises:

[0149] The face acquisition module 601 is used to execute 6011. The platform enters the face information of the students being assessed through the face acquisition terminal to form an evaluation management face information database.

[0150] The facial recognition module 602 is used to execute 6021, accurately identifying the identification information of the student currently being assessed and indexing the target assessment items. For example, if a teacher sees a student behaving uncivilly in a teaching scenario, they can use the facial recognition module built into a mobile app to identify the student's information and index the target assessment items.

[0151] The speech recognition module 603 is used to execute 6031, recognize the teacher's language evaluation in real time, convert it into the system's evaluation points, analyze the scores or deductions of the evaluated students, and make a judgment.

[0152] In some embodiments, 6011, 6021, and 6031 may be executed sequentially.

[0153] In this embodiment, firstly, speech recognition technology is used to quickly convert teachers' language skills into effective text information, improving the efficiency of information input for teachers' moral education evaluation management. Secondly, facial recognition technology expands the evaluation process to any teaching scenario, allowing teachers to accurately locate the evaluated student in any location, such as the classroom or playground, and make real-time, recorded evaluations. In traditional school management scenarios, teachers' effective management radius is limited to students in their own classes. The application of AI facial recognition technology allows any teacher in the school to quickly identify any student in any class. This technology expands the management radius and makes student moral education evaluation data more objective and accurate.

[0154] Based on the foregoing embodiments, this application provides a scoring device. The modules and units included in the device can be implemented by a processor in a computer device; of course, they can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), a microprocessor unit (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0155] This application embodiment further provides a scoring device, Figure 7 This is a schematic diagram of the composition structure of the scoring device provided in the embodiments of this application, as shown below. Figure 7 As shown, the scoring device 700 includes:

[0156] The first acquisition module 701 is used to acquire the identification information of the assessment object and the voice evaluation information of the assessment object, wherein the voice evaluation information includes at least the behavioral characteristic information of the assessment object.

[0157] The second acquisition module 702 is used to acquire the trained scoring model;

[0158] The scoring module 703 is used to process the speech evaluation information using the trained scoring model to obtain the assessment target of the assessment object and the scoring result corresponding to the assessment target.

[0159] The output module 704 is used to output the identification information of the assessment object, the assessment target, and the scoring result corresponding to the assessment target.

[0160] In some embodiments, the scoring module 703 is further configured to use the preset scoring model to score the sample speech evaluation information, thereby obtaining the predicted assessment target corresponding to the sample speech evaluation information and the predicted score result of the predicted assessment target; the scoring device 700 further includes:

[0161] The third acquisition module is used to acquire a preset scoring model, sample voice evaluation information, sample assessment targets corresponding to the sample voice evaluation information, and sample scoring results corresponding to the sample assessment targets.

[0162] The first determining module is used to determine that when the sample assessment target and the predicted assessment target meet the preset matching conditions, it obtains the error information between the sample scoring result and the predicted scoring result.

[0163] The first training module is used to perform backpropagation training on the preset scoring model based on the error information and the preset error threshold to obtain the trained scoring model.

[0164] In some embodiments, the scoring device 700 further includes:

[0165] The second determining module is used to determine the set of reference assessment targets corresponding to the assessment object based on the identification information of the assessment object;

[0166] The third determining module is used to determine the reference assessment target associated with the voice evaluation information from the reference assessment target set when the assessment target is not included in the reference assessment target set.

[0167] The second training module is used to continue training the trained scoring model based on the voice evaluation information and the reference assessment target to obtain a retrained scoring model.

[0168] In some embodiments, the scoring device 700 further includes:

[0169] The decoding module is used to decode the speech evaluation information to obtain the decoded speech evaluation information;

[0170] The sampling module is used to sample the decoded speech evaluation information to obtain sampled speech evaluation information;

[0171] The recognition module is used to perform speech recognition on the sampled speech evaluation information using a preset acoustic model to obtain text data;

[0172] The input module is used to input the text data into the trained scoring model to obtain the assessment target of the assessment object and the scoring result corresponding to the assessment target.

[0173] In some embodiments, the input module includes:

[0174] The first determining unit is used to input the text data into the trained scoring model and determine the set of similar assessment targets of the text data based on the text data and the first preset search algorithm.

[0175] The second determining unit is used to determine the assessment target of the text data based on the set of similar assessment targets and the second preset search algorithm;

[0176] The third determining unit is used to determine the assessment target of the text data as the assessment target of the assessment object.

[0177] In some embodiments, the scoring device 700 further includes:

[0178] The fourth acquisition module is used to acquire the image of the assessment subject to be recognized;

[0179] An extraction module is used to determine a facial image from the image to be identified and to extract facial feature information from the facial image;

[0180] The fourth determining module is used to determine the identification information of the assessment subject from a preset information database based on the facial feature information, wherein the preset information database stores the correspondence between facial feature information and identification information.

[0181] In some embodiments, the output module is further configured to output a prompt message when it is determined that the identification information of the assessment subject cannot be identified from a preset information database, the prompt message being used to prompt the input of the identification information of the assessment subject; the scoring device 700 further includes:

[0182] The response module is used to respond to the received input operation and obtain the identification information of the assessment object;

[0183] A module is established to establish the correspondence between the facial feature information of the assessment subject and the identification information of the assessment subject;

[0184] A storage module is used to store the correspondence into the preset information database.

[0185] It should be noted that the description of the scoring device in this application embodiment is similar to the description of the method embodiment above, and has similar beneficial effects as the method embodiment; therefore, it will not be repeated. For technical details not disclosed in this device embodiment, please refer to the description of the method embodiment in this application for understanding.

[0186] It should be noted that, in the embodiments of this application, if the above-described scoring method is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or the part that contributes to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware and software combination.

[0187] Accordingly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the scoring method provided in the above embodiments.

[0188] This application provides a scoring device. Figure 8 This is a schematic diagram of the composition structure of the scoring device provided in the embodiments of this application, such as... Figure 8 As shown, the scoring device 800 includes: a processor 801, at least one communication bus 802, a user interface 803, at least one external communication interface 804, and a memory 805. The communication bus 802 is configured to enable communication between these components. The user interface 803 may include a display screen, and the external communication interface 804 may include standard wired and wireless interfaces. The processor 801 is configured to execute a program of a scoring method stored in the memory to implement the steps of the scoring method provided in the above embodiment.

[0189] The descriptions of the scoring device and storage medium embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. For technical details not disclosed in the scoring device and storage medium embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0190] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above-described processes do not imply a sequential order of execution; the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above-described embodiments are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0191] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0192] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0193] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0194] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0195] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0196] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an AC to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROMs, magnetic disks, or optical disks.

[0197] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A scoring method, characterized in that, The method includes: Obtain the identification information of the assessment subject and the voice evaluation information of the assessment subject, wherein the voice evaluation information includes at least the behavioral characteristic information of the assessment subject; Obtain the trained scoring model and determine the assessment objectives for the assessment subjects; When the set of reference assessment targets determined based on the identification information does not include the assessment target, the similarity between each reference assessment target in the set of reference assessment targets and the voice evaluation information is calculated, and the reference assessment target with the highest similarity is determined as the reference assessment target associated with the voice evaluation information. Based on the voice evaluation information and the reference assessment target, the trained scoring model is trained again to obtain a retrained scoring model; The trained scoring model or the retrained scoring model is used to score the speech evaluation information to obtain the assessment target of the assessment object and the scoring result corresponding to the assessment target. Output the identification information of the assessment object, the assessment target, and the scoring result corresponding to the assessment target.

2. The method according to claim 1, characterized in that, The method further includes: Obtain a preset scoring model, sample speech evaluation information, sample assessment targets corresponding to the sample speech evaluation information, and sample scoring results corresponding to the sample assessment targets; The sample speech evaluation information is scored using the preset scoring model to obtain the predicted assessment target and the predicted score result of the predicted assessment target corresponding to the sample speech evaluation information. When it is determined that the sample assessment target and the predicted assessment target meet the preset matching conditions, the error information between the sample scoring result and the predicted scoring result is obtained; The preset scoring model is trained by backpropagation based on the error information and the preset error threshold to obtain the trained scoring model.

3. The method according to claim 1, characterized in that, The method further includes: Based on the identification information of the assessment object, determine the set of reference assessment targets corresponding to the assessment object; When the set of reference assessment targets includes the assessment target, the trained scoring model remains unchanged.

4. The method according to claim 1, characterized in that, The method further includes: The speech evaluation information is decoded to obtain the decoded speech evaluation information; The decoded speech evaluation information is sampled to obtain sampled speech evaluation information; The sampled speech evaluation information is subjected to speech recognition using a preset acoustic model to obtain text data. Accordingly, the step of using the trained scoring model or the retrained scoring model to score the speech evaluation information, and obtaining the assessment target of the assessment object and the scoring result corresponding to the assessment target, includes: The text data is input into the trained scoring model or the retrained scoring model to obtain the assessment target of the assessment object and the scoring result corresponding to the assessment target.

5. The method according to claim 4, characterized in that, The text data is input into the trained scoring model or the retrained scoring model to obtain the assessment objectives for the assessment subject, including: The text data is input into the trained scoring model or the retrained scoring model, and a set of similar assessment targets for the text data is determined based on the text data and the first preset search algorithm. The assessment objectives of the text data are determined based on the set of similar assessment objectives and the second preset search algorithm. The assessment objectives of the text data are determined as the assessment objectives of the assessment object.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: Obtain the image of the assessment object to be identified; A facial image is determined from the image to be identified, and facial feature information of the facial image is extracted; Based on the facial feature information, the identification information of the assessment subject is determined from a preset information database, wherein the preset information database stores the correspondence between facial feature information and identification information.

7. The method according to claim 6, characterized in that, The method further includes: When it is determined that the identification information of the assessment object cannot be identified from the preset information database, a prompt message is output, which is used to prompt the input of the identification information of the assessment object; In response to the received input operation, the identification information of the assessment object is obtained; Establish a correspondence between the facial feature information of the assessment subject and the identification information of the assessment subject; The correspondence is stored in the preset information database.

8. A scoring device, characterized in that, The device includes: The first acquisition module is used to acquire the identification information of the assessment object and the voice evaluation information of the assessment object, wherein the voice evaluation information includes at least the behavioral characteristic information of the assessment object. The second acquisition module is used to acquire the trained scoring model and determine the assessment target of the assessment object; when the assessment target is not included in the reference assessment target set determined based on the identification information, the similarity between each reference assessment target in the reference assessment target set and the speech evaluation information is calculated, and the reference assessment target with the highest similarity is determined as the reference assessment target associated with the speech evaluation information; the trained scoring model is further trained based on the speech evaluation information and the reference assessment target to obtain a retrained scoring model; The scoring module is used to score the speech evaluation information using the trained scoring model or the retrained scoring model, so as to obtain the assessment target of the assessment object and the scoring result corresponding to the assessment target. The output module is used to output the identification information of the assessment object, the assessment target, and the scoring result corresponding to the assessment target.

9. A scoring device, characterized in that, The device includes: Processor; and Memory for storing computer programs that can run on the processor; When the computer program is executed by a processor, it implements the scoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions configured to perform the scoring method according to any one of claims 1 to 7.

11. A computer software product comprising computer-executable instructions, characterized in that, The computer executes the instructions to implement the method of any one of claims 1 to 7.

Images