Answer content marking method, system and related devices

Abstract

The present application discloses a method, a system and related devices for correcting answer content. The method includes: obtaining a target image corresponding to the current answer sheet, and using an image recognition large model to determine at least one target partition corresponding to the target image; generating recognition prompt information matching the target partition based on the position information of the target partition in the target image, inputting the recognition prompt information into the recognition large model, and using the prompt information to generate a recognition result matching the target image; generating a target correction result for obtaining the answer content matching the recognition result based on the recognition result. By the above method, the present application can improve the efficiency and accuracy of correcting answer content.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular, to a method, system, and related device for correcting answer content. Background Art

[0002] In the field of education, the correction of answer content is an important part of the teaching process. There are many drawbacks in the traditional manual correction method. With the continuous expansion of the education scale and the in-depth development of education informatization, automatic correction technology has emerged. The current automatic correction technology mainly relies on OCR recognition technology to recognize the answer sheets and determines whether the corresponding answer content is answered correctly according to the recognized content. Due to handwriting differences, this method is prone to errors during recognition, resulting in poor correction effects.

[0003] In view of this, how to propose an efficient and accurate method for correcting answer content has become an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem to be solved by this application is to provide a method, system, and related device for correcting answer content, which can improve the efficiency and accuracy of correcting answer content.

[0005] To solve the above technical problem, a technical solution adopted by this application is: to provide a method for correcting answer content, including: obtaining a target image corresponding to the current answer sheet, and using an image recognition large model to determine at least one target partition corresponding to the target image; based on the position information of the target partition in the target image, generating recognition prompt information that matches the target partition, inputting the recognition prompt information into the recognition large model, and using the prompt information to generate a recognition result that matches the target image; based on the recognition result, generating a target correction result that matches the answer content in the recognition result.

[0006] To solve the above technical problems, another technical solution adopted by this application is: to provide an answer content marking system, including: an acquisition module, configured to acquire a target image corresponding to the current answer sheet, and use an identification large model to determine at least one target partition corresponding to the target image; an identification module, configured to generate identification prompt information matching the target partition based on the position information of the target partition in the target image, input the identification prompt information into the identification large model, and generate an identification result matching the target image; a processing module, configured to generate a target marking result matching the answer content in the identification result based on the identification result. To solve the above technical problems, another technical solution adopted by this application is: to provide an electronic device, including: a memory and a processor coupled to each other, where program instructions are stored in the memory, and the processor is configured to execute the program instructions to implement the answer content marking method mentioned in the above technical solution.

[0007] To solve the above technical problems, another technical solution adopted by this application is: to provide a computer-readable storage medium, on which program instructions are stored, and when the program instructions are executed by a processor, the answer content marking method mentioned in the above technical solution is implemented.

[0008] The beneficial effect of this application is: different from the prior art, the answer content marking method proposed by this application detects at least one corresponding target partition from the target image corresponding to the current answer sheet by using an identification large model, and generates corresponding identification prompt information according to the position information of the target partition in the target image. Using the semantic information of the identification large model combined with the identification prompt information, an identification result matching the target image is identified from the target image, improving the accuracy of obtaining the identification result. Furthermore, the answer content is automatically marked according to the identification result to improve the marking efficiency. Description of the Drawings

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings. Among them:

[0010] Figure 1 It is a schematic flowchart of an implementation manner of the answer content marking method of this application;

[0011] Figure 2 is Figure 1 The schematic flowchart of another implementation manner corresponding to step S101 in

[0012] Figure 3 is Figure 1The flowchart corresponding to step S102 in another embodiment;

[0013] Figure 4 is Figure 1 The flowchart corresponding to step S102 in yet another embodiment;

[0014] Figure 5 is Figure 1 The flowchart corresponding to step S103 in another embodiment;

[0015] Figure 6 is the structural diagram corresponding to an embodiment of the answer content marking system of the present application;

[0016] Figure 7 is the structural diagram corresponding to an embodiment of the electronic device of the present application;

[0017] Figure 8 is the structural diagram corresponding to an embodiment of the computer-readable storage medium of the present application. Specific Embodiments

[0018] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments, and different embodiments can be adaptively combined. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0019] Please refer to Figure 1 , Figure 1 is the flowchart of an embodiment of the answer content marking method of the present application. The method includes:

[0020] S101: Obtain the target image corresponding to the current answer sheet, and use the recognition large model to determine at least one target partition corresponding to the target image.

[0021] In one embodiment, image acquisition is performed on the current answer sheet to be marked to obtain the corresponding target image. The recognition large model is used to perform region detection on the target image to obtain at least one target partition corresponding to the target image.

[0022] Specifically, the recognition large model is used to perform region detection on the target image to obtain at least one corresponding target partition, and the position information corresponding to each target partition is determined. Among them, the target partition is at least one boxed area determined according to the layout structure of the content in the target image.

[0023] In an implementation scenario, the above-mentioned recognition large model is a large language model with relatively excellent data analysis capabilities. By inputting the target image into the recognition large model and prompting the recognition large model to perform partition detection, the target partition output by the recognition large model can be obtained.

[0024] In a specific application scenario, the above-mentioned large language model may include, but is not limited to, Deep Neural Networks (DNNs), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM), and Generative Pretrained Transformer models, etc. There is no specific limitation on the specific structure and specific deployment of the large language model here.

[0025] S102: Based on the position information of the target partition in the target image, generate recognition prompt information matching the target partition, input the recognition prompt information into the recognition large model, and generate a recognition result matching the target image.

[0026] In an implementation manner, according to the position information corresponding to each target partition, generate matching recognition prompt information. Input the recognition prompt information into the recognition large model, and use the recognition large model to perform semantic analysis on the recognition prompt information to obtain a recognition result from the target image. Among them, the above-mentioned recognition result includes at least one of stem information and answer content, etc.

[0027] S103: Based on the recognition result, generate a target marking result matching the answer content in the recognition result.

[0028] In an implementation manner, after obtaining the recognition result, obtain the reference answer matching the answer content in the recognition result. According to the mutually matching reference answer and answer content, obtain the corresponding target marking result.

[0029] Specifically, determine whether the answer content is consistent with the matching reference answer. If it is consistent, the generated target marking result indicates that the corresponding answer content is answered correctly. If it is not consistent, the generated target marking result indicates that the corresponding answer content is answered incorrectly.

[0030] In another implementation manner, for the answer content in the recognition result, determine at least one candidate character corresponding to the recognized character according to the recognized character in the answer content. Replace the corresponding recognized character with the candidate character to obtain the reference content corresponding to the answer content. Obtain the reference answer matching the answer content, and according to the reference answer, answer content, and the corresponding reference content, obtain the corresponding target marking result.

[0031] Specifically, the candidate characters are other characters similar to the corresponding recognized characters. By using the candidate characters to replace the recognized characters, the corresponding reference content is obtained, so as to reduce the impact of recognition errors caused by handwriting habits on the target marking result and improve the flexibility of marking.

[0032] In a specific application scenario, when the handwritten answer content in the recognition result includes "stard", to avoid the impact of recognition errors caused by handwriting habits on the marking result, some recognized characters in "stard" are replaced to obtain the corresponding reference content "stand". The answer content "stard" and the reference content "stand" are respectively compared with the corresponding reference answer. And when any one of the answer content and the reference content is consistent with the reference answer, it is determined that the target marking result of the corresponding question is correct. Otherwise, it is determined that the target marking result of the corresponding question is wrong.

[0033] In another implementation manner, the implementation process of step S103 may further include: obtaining a reference answer that matches the answer content. Based on the answer content and the reference answer, generating marking prompt information, and inputting the marking prompt information into the intelligent analysis model to generate a target marking result that matches the answer content.

[0034] Specifically, obtain a preset marking prompt template, input the answer content and the reference answer into the corresponding positions in the marking prompt template to obtain the marking prompt information. Input the marking prompt information into the intelligent analysis model, use the intelligent analysis model to perform semantic recognition on the marking prompt information, and generate a target marking result. For example, the preset marking prompt template is "The standard reference answer is: <stdanswer>, the user's answer is: <useranswer>"Give the target marking result corresponding to the answer content", and by inputting the answer content and the reference answer into the corresponding areas of the marking prompt template, marking prompt information is obtained.

[0035] In one implementation scenario, the above intelligent analysis model is also a large language model with relatively excellent data analysis capabilities. By inputting the marking prompt information into the intelligent analysis model, the intelligent analysis model generates the target marking result according to the semantic information corresponding to each answer content and the reference answer respectively.

[0036] In another implementation scenario, the above intelligent analysis model is the same model as the recognition large model mentioned in the above corresponding implementation manner. So that after obtaining the target image, the target image is input into the recognition large model, and finally the recognition large model outputs the corresponding target marking result, greatly improving the marking efficiency.

[0037] The answer content marking method proposed in this application, after using the recognition large model to detect at least one corresponding target partition from the target image corresponding to the current answer sheet, generates the corresponding recognition prompt information according to the position information of the target partition in the target image. Using the recognition large model to combine the semantic information of the recognition prompt information to identify the recognition result that matches the target image from the target image improves the accuracy of obtaining the recognition result. Furthermore, the answer content is automatically marked according to the recognition result to improve the marking efficiency.

[0038] Please refer to Figure 2 , Figure 2 is Figure 1 The flowchart corresponding to step S101 in another implementation manner. Specifically, the implementation process of step S101 includes:

[0039] S201: Collect the current answer sheet to obtain the target image.

[0040] In one implementation manner, after obtaining the current answer sheet, use the collection device to collect an image of the current answer sheet to obtain the corresponding target image.

[0041] In one implementation scenario, the above collection device includes a camera. By setting the current answer sheet in the corresponding position and triggering the collection instruction, the camera on the collection device is used to collect the target image corresponding to the current answer sheet.

[0042] In another implementation manner, obtain the initial image collected from the current answer sheet. Among them, the initial image is matched with the corresponding initial length and initial width.

[0043] Specifically, use the collection device to collect the current answer sheet to obtain the corresponding initial image, and obtain the initial length and initial width corresponding to the initial image. Among them, the initial length is greater than or equal to the initial width.

[0044] Further, obtain a reference correction length matching the initial length and a reference correction width matching the initial width, and based on the reference correction length and the reference correction width, adjust the size of the initial image to obtain an adjusted target image.

[0045] Specifically, obtain a preset reference correction length and a reference correction width. Compare the initial width of the initial image with the reference correction width. In response to the initial width being greater than the reference correction width, determine a first correction coefficient, and use the first correction coefficient to correct the initial length and the initial width of the initial image, that is, multiply the initial length and the initial width of the initial image by the first correction coefficient respectively to obtain a first corrected image, such that the width of the first corrected image is equal to the reference correction width. Further, compare the length of the first corrected image with the reference correction width. In response to the length of the first corrected image being greater than the reference correction length, determine a second correction coefficient, and use the second correction coefficient to correct the length and the width of the first corrected image, that is, multiply the length and the width of the first corrected image by the second correction coefficient respectively to obtain a target image, such that the length of the target image is equal to the reference correction length. By adjusting the size of the initial image using the determined reference correction length and reference correction width, it is convenient to perform unified processing on the target images of different answer sheets.

[0046] It should be noted that in the above correction process, if the initial width of the initial image is less than or equal to the reference correction width, no preliminary adjustment is made to the initial image, that is, the initial image is used as the first corrected image. And if the length of the first corrected image is less than or equal to the reference correction length, the first corrected image is directly used as the target image.

[0047] In another embodiment, to improve the image quality of the initial image, after obtaining the initial image, preprocess the initial image.

[0048] Specifically, the above preprocessing process includes performing morphological processing on the initial image to remove noise in the initial image. Alternatively, the preprocessing process further includes performing bilinear interpolation processing on the initial image to improve the smoothness of the initial image.

[0049] S202: Use the trained region detection module to obtain the image coding features corresponding to the target image.

[0050] In one embodiment, obtain a trained region detection module, and input the target image corresponding to the current answer sheet into the region detection module to encode the target image using the region detection module to obtain the image coding features corresponding to the target image. By obtaining the image coding features, it is convenient for the subsequent recognition large model to analyze the target image.

[0051] In an implementation scenario, the above-mentioned region detection module is trained using multiple first training samples, so that the trained region detection module has the ability to extract image features from the target image.

[0052] S203: Input the image encoding features into the recognition large model to generate at least one target partition corresponding to the target image; wherein, the target partition is matched with corresponding position information, and the position information includes the reference coordinates corresponding to the target partition.

[0053] In an implementation manner, according to the image encoding features, corresponding partition prompt information is generated, the partition prompt information is input into the recognition large model, and after the recognition large model analyzes the partition prompt information, at least one target partition matching the target image and the position information corresponding to each target partition are generated.

[0054] Specifically, obtain the pre-constructed partition prompt template. After obtaining the image encoding features, input the image encoding features into the corresponding position in the partition prompt template to generate the corresponding partition prompt information, which is used to prompt the recognition large model to detect at least one target partition from the target image according to the image encoding features.

[0055] In a specific application scenario, the partition prompt template is "According to the <image encoding features> of the target image, perform partition detection on the target partition, and output at least one corresponding target partition and the position information of each target partition". Use the image encoding features corresponding to the target image to replace "<image encoding features>" in the partition prompt template to obtain the partition prompt information.

[0056] In another implementation manner, before using the recognition large model to determine the target partition, use multiple second training samples to fine-tune the recognition large model to optimize the effect of the fine-tuned recognition large model in performing partition detection. Among them, each second training sample is matched with a corresponding training label, and the training label includes the pre-annotated boxed area and the coordinates of the four vertices of each boxed area.

[0057] The above solution helps to subsequently obtain the recognition result in the target image by combining the target partition recognition by using the recognition large model to output the corresponding target partition according to the partition prompt information, and reduces the probability of content interference between different questions.

[0058] In another implementation manner, the recognition large model includes a partition module. Based on this, step S101 includes: obtaining the trained partition module, inputting the target image corresponding to the current answer sheet into the partition module, so as to use the partition module to detect at least one target partition in the target image and determine the position information corresponding to each target partition.

[0059] Specifically, the above-mentioned trained partitioning module includes an image encoding sub-module and an image decoding sub-module that are coupled to each other. After inputting the target image into the partitioning module, the image encoding sub-module in the partitioning module processes the target image to obtain the image encoding features corresponding to the target image. The image encoding features are input into the image decoding sub-module to analyze the layout structure of the content in the target image by using the image decoding sub-module, so as to output the target partitions in the target image and the corresponding position information of each target partition.

[0060] In an implementation scenario, the above-mentioned partitioning module is trained by using a plurality of third training samples. Each third training sample is matched with a corresponding training label, and the training label includes a pre-annotated boxed area and the coordinates of the four vertices of each boxed area. By using a plurality of third training samples to train the partitioning module, the trained partitioning module is enabled to have the ability to detect the corresponding target partitions from the target image. Among them, the specific structure of the partitioning module can refer to the existing neural network model structure and will not be elaborated in detail here.

[0061] Please refer to Figure 3 , Figure 3 is Figure 1 the schematic flowchart of another implementation manner corresponding to step S102 in

[0062] S301: Generate recognition prompt information matching the target partition based on the position information corresponding to the target partition.

[0063] In an implementation manner, recognition prompt information is generated according to the reference coordinates corresponding to the target partition.

[0064] In a specific application scenario, a pre-constructed recognition prompt template is obtained. For each target partition, the corresponding reference coordinates are filled into the corresponding position in the recognition prompt template to obtain the corresponding recognition prompt information. For example, if the pre-constructed recognition prompt template is "Detect the content of the target partition according to the <reference position> of the target partition and output the corresponding recognition result", then when the reference coordinates corresponding to target partition A are , , and , the above four reference coordinates are used to replace the "<reference position>" in the recognition prompt template to obtain the prompt information corresponding to target partition A.

[0065] In another implementation manner, in response to obtaining the image encoding features corresponding to the target image in the above corresponding implementation manner, to improve the efficiency of obtaining the recognition result, the implementation process of the above step S301 can also be: Generate recognition prompt information according to the image encoding features corresponding to the target image and the reference coordinates corresponding to the target partition.

[0066] In a specific application scenario, obtain a pre-constructed recognition prompt template. For each target partition, fill in the corresponding reference coordinates at the corresponding positions in the recognition prompt template to obtain the corresponding recognition prompt information. For example, if the pre-constructed recognition prompt template is "Perform content detection on the target partition according to the <image coding feature> and the <reference position> of the target partition, and output the corresponding recognition result", then fill in the image coding feature of the target image and the reference coordinates of the target partition at the corresponding positions in the recognition prompt template to obtain the recognition prompt information.

[0067] S302: Input the recognition prompt information into the recognition large model, and use the recognition large model to generate a structured recognition result that matches the target partition; wherein, the recognition result includes recognition content that matches at least one recognition category.

[0068] In one embodiment, input the recognition prompt information into the recognition large model, so that after the recognition large model analyzes the recognition prompt information, a matching structured recognition result is recognized from the target image. The above recognition result includes recognition content that matches at least one recognition category. Among them, the recognition categories include printed body category, handwritten category, answer area category, image category, table category, etc.

[0069] Among them, the recognition result recognized by the recognition large model includes recognition content that matches the recognition category, so that the recognition large model distinguishes the answer content in the recognition content according to the recognition category. For example, use the recognition content corresponding to the handwritten category as the answer content, or use the recognition content corresponding to the answer area category as the answer content.

[0070] In the above solution, by inputting the recognition prompt information into the recognition large model, the recognition large model combines the positional association relationship between different target partitions and the semantic information of the content in each target partition to recognize the recognition results that respectively match each target partition, reducing the probability of recognizing the content in other target partitions as the recognition result corresponding to the current target partition, and improving the recognition accuracy.

[0071] Please refer to Figure 4 , Figure 4 is Figure 1 a schematic flowchart of another embodiment corresponding to step S102 in

[0072] S401: Obtain a target recognition mode that matches the target image.

[0073] In one embodiment, pre-determine a target recognition mode that matches the target image, and this target recognition mode is used to represent the output mode of the recognition result.

[0074] Specifically, the target recognition mode includes a small-question output mode, a large-question output mode, etc. When the target recognition mode is the small-question output mode, when performing content recognition on the target image subsequently, the corresponding recognition results are output in units of each small question. Or, when the target recognition mode is the large-question output mode, subsequently, the corresponding recognition results are output in units of each large question.

[0075] In an implementation scenario, when the target recognition mode is the large-question output mode, when outputting the recognition result corresponding to the current large question, the association information between each small question in the current large question is combined to improve the accuracy of the recognition result. And when the target recognition mode is the small-question output mode, when outputting the recognition result corresponding to the current large question, there is no need to combine the association information between different small questions, which improves the efficiency of obtaining the recognition result. Among them, the above-mentioned association information includes semantic association information and position association information.

[0076] Optionally, in other implementation manners, the target recognition mode further includes a line output mode, etc., that is, when the current target recognition mode is the line output mode, when performing content recognition on the target image, the corresponding recognition results are output in units of each line.

[0077] S402: Generate recognition prompt information based on the target recognition mode and the position information corresponding to the target partition.

[0078] In an implementation manner, recognition prompt information is generated according to the target recognition mode, the reference coordinates corresponding to each target partition, and the target image. The recognition prompt information is used to prompt the recognition large model to recognize the recognition result corresponding to the target image according to the target recognition mode and the detected reference coordinates corresponding to each target partition.

[0079] S403: Input the recognition prompt information into the recognition large model to generate a structured recognition result matching the target image; among them, the recognition result includes recognition content matching at least one recognition category.

[0080] In an implementation manner, the recognition prompt information is input into the recognition large model so that after the recognition large model analyzes the recognition prompt information, a matching structured recognition result is recognized from the target image. The above recognition result includes recognition content matching at least one recognition category. Among them, the recognition categories include printed body category, handwritten category, answer area category, image category, table category, etc., and the specific implementation process can refer to the above corresponding implementation manners.

[0081] The above solution helps to improve the flexibility of obtaining recognition results by obtaining the target recognition mode and generating recognition prompt information in combination with the target recognition mode.

[0082] Please refer to Figure 5 , Figure 5 Yes Figure 1 It is a schematic flowchart of step S103 corresponding to another implementation manner. Specifically, the implementation process of step S103 includes:

[0083] S501: Obtain the reference answer that matches the answer content. Based on the reference answer, obtain the initial marking result that matches the answer content.

[0084] In one implementation manner, after obtaining the recognition result, when the recognition result includes the answer content, obtain the reference answer that matches the answer content. According to the mutually matching reference answer and answer content, obtain the corresponding initial marking result.

[0085] Specifically, when the answer content is consistent with the corresponding reference answer, generate the corresponding initial marking result as "correct". Otherwise, generate the corresponding initial marking result as "wrong" and deduct the corresponding score.

[0086] In another implementation manner, when the recognition result includes the answer content, according to the recognized characters in the answer content, determine at least one candidate character corresponding to the recognized character. Use the candidate character to replace the corresponding recognized character to obtain the reference content corresponding to the answer content. Obtain the reference answer that matches the answer content. According to the reference answer, answer content, and the corresponding reference content, obtain the corresponding initial marking result. The specific implementation process can refer to the above corresponding implementation manner and will not be elaborated in detail here.

[0087] S502: Based on the initial marking result, obtain the shunt judgment result.

[0088] In one implementation manner, when the initial marking result is "correct", generate a shunt judgment result to determine that the initial marking result meets the preset conditions. Or, when the initial marking result is "wrong", generate a shunt judgment result to determine that the initial marking result does not meet the preset conditions.

[0089] S503: In response to the shunt judgment result determining that the initial marking result meets the preset conditions, use the initial marking result as the target marking result.

[0090] In one implementation manner, when it is determined that the initial marking result meets the preset conditions, directly use the initial marking result as the target marking result.

[0091] S504: In response to the shunt judgment result determining that the initial marking result does not meet the preset conditions, generate a marking prompt message based on the answer content and reference answer, input the marking prompt message into the intelligent analysis model, and generate the target marking result that matches the answer content.

[0092] In one embodiment, when it is determined that the initial grading result does not meet the preset conditions, the answer content and the reference answer are input into the intelligent analysis model to semantically analyze the answer content and the reference answer respectively by using the intelligent analysis model, and a target grading result matching the answer content is obtained according to the semantics.

[0093] In one implementation scenario, a preset grading prompt template is obtained, the answer content and the reference answer are input into the corresponding positions in the grading prompt template to obtain grading prompt information. The grading prompt information is input into the intelligent analysis model, and the intelligent analysis model is used to perform semantic recognition on the grading prompt information and generate a target grading result.

[0094] In a specific application scenario, the pre-constructed task template is "The stem information is: <content>, the standard reference answer is: <stdanswer>, the user's answer content is: <useranswer>"Give the target marking result corresponding to the answer content". By inputting the answer content, the question information corresponding to the answer content, and the reference answer into the corresponding areas of the task template, the task text is obtained.

[0095] In another embodiment, to improve the accuracy of marking, the process of using the intelligent analysis model to generate the target marking result in the above corresponding embodiment can be: using the intelligent analysis model to obtain the matching target marking result according to the answer content, the question information matched by the answer content, the reference answer matched by the answer content, and the corresponding problem-solving process of the reference answer.

[0096] Specifically, prompt the intelligent analysis model to generate a problem-solving process that matches the above reference answer, and combine the problem-solving process, the answer content, the question information corresponding to the answer content, and the reference answer corresponding to the answer content to generate the target marking result.

[0097] In a specific application scenario, the pre-constructed task template is "Complete the following questions and give the problem-solving process. The question information is: <content>, the standard reference answer is: <stdanswer>, the user's answer content is: <useranswer>, combined with the problem-solving process, give the target marking result corresponding to the answer content. By inputting the answer content, the question stem information matched with the answer content, and the reference answer into the corresponding areas of the task template, the task text is obtained.

[0098] For the initial marking result that does not meet the preset conditions, the above solution calls the intelligent analysis model to obtain the target marking result at least based on the semantics of the answer content and the reference answer, reducing the probability of marking errors.

[0099] In another embodiment, the implementation process of the above step S502 may further include: obtaining feature information matching the answer content. Among them, the above feature information includes at least one of the question stem information matching the answer content, the number of characters corresponding to the reference answer, and the language category.

[0100] Specifically, obtain the question stem information matching the answer content and the reference answer matching the answer content, determine the number of characters corresponding to the reference answer and the language category corresponding to the reference answer, and use at least one of the answer content, the matching question stem information, the matching reference answer, the number of characters corresponding to the reference answer, and the language category corresponding to the reference answer as feature information. Among them, the above language category includes Chinese, English, Japanese, etc.

[0101] Optionally, when the language category corresponding to the reference answer is English, the feature information may further include the character case information corresponding to the reference answer.

[0102] Further, input the initial marking result and the feature information into the trained shunt judgment model to obtain the shunt judgment result output by the shunt judgment model.

[0103] Specifically, input the initial marking result and the above feature information into the trained shunt judgment model to analyze the initial marking result and the feature information by using the shunt judgment model and output the shunt judgment result.

[0104] In an implementation scenario, the above shunt judgment model is a pre-trained language model, which is trained by using a plurality of second training samples. The specific structure of the shunt judgment model can refer to the existing BERT model structure.

[0105] The above solution automatically judges the credibility of the initial marking result by training the shunt judgment model to decide whether to call the intelligent analysis model for further marking, improving the accuracy of judging the answer content.

[0106] Please refer to Figure 6 , Figure 6 It is a schematic structural diagram corresponding to an implementation manner of the answer content marking system of the present application. Specifically, the answer content marking system includes an acquisition module 10, an identification module 20, and a processing module 30 that are mutually coupled.

[0107] Specifically, the acquisition module 10 is used to acquire a target image corresponding to the current answer sheet, and use an identification large model to determine at least one target partition corresponding to the target image.

[0108] The identification module 20 is used to generate identification prompt information matching the target partition based on the position information of the target partition in the target image, input the identification prompt information into the identification large model, and generate an identification result matching the target image.

[0109] The processing module 30 is used to generate a target marking result matching the answer content in the identification result based on the identification result.

[0110] In one implementation manner, the acquisition module 10 acquires a target image corresponding to the current answer sheet, and uses an identification large model to determine at least one target partition corresponding to the target image, including: collecting the current answer sheet to obtain a target image; using a trained region detection module to obtain the image coding features corresponding to the target image; inputting the image coding features into the identification large model to generate at least one target partition corresponding to the target image; wherein, the target partition is matched with corresponding position information, and the position information includes the reference coordinates corresponding to the target partition.

[0111] In one implementation manner, please continue to refer to Figure 6 , the answer content marking system proposed by the present application further includes an image adjustment module 40 coupled to the acquisition module 10, and the image adjustment module 40 is used to: acquire an initial image obtained by collecting the current answer sheet; wherein, the initial image is matched with corresponding initial length and initial width; acquire a reference correction length matching the initial length and a reference correction width matching the initial width, and based on the reference correction length and the reference correction width, adjust the size of the initial image to obtain an adjusted target image.

[0112] In one implementation manner, the identification module 20 generates identification prompt information matching the target partition based on the position information of the target partition in the target image, inputs the identification prompt information into the identification large model, and generates an identification result matching the target image, including: acquiring a target identification mode matching the target image; generating identification prompt information based on the target identification mode and the position information corresponding to the target partition; inputting the identification prompt information into the identification large model to generate a structured identification result matching the target image; wherein, the identification result includes identification content matching at least one identification category.

[0113] In one embodiment, the processing module 30 generates a target marking result that matches the answer content in the recognition result, including: obtaining a reference answer that matches the answer content; generating a marking prompt message based on the answer content and the reference answer, and inputting the marking prompt message into the intelligent analysis model to generate a target marking result that matches the answer content.

[0114] In one embodiment, the processing module 30 generates a target marking result that matches the answer content in the recognition result, including: obtaining a reference answer that matches the answer content, and obtaining an initial marking result that matches the answer content based on the reference answer; obtaining a diversion judgment result based on the initial marking result; in response to the diversion judgment result determining that the initial marking result meets the preset conditions, using the initial marking result as the target marking result; in response to the diversion judgment result determining that the initial marking result does not meet the preset conditions, generating a marking prompt message based on the answer content and the reference answer, and inputting the marking prompt message into the intelligent analysis model to generate a target marking result that matches the answer content.

[0115] In one embodiment, please continue to refer to Figure 6 , the answer content marking system proposed in this application further includes a diversion module 50 coupled to the processing module 30. The diversion module 50 is used to obtain feature information that matches the answer content; wherein, the feature information includes at least one of the stem information that matches the answer content, the number of characters corresponding to the reference answer, and the language category; inputting the initial marking result and the feature information into the trained diversion judgment model to obtain the diversion judgment result output by the diversion judgment model.

[0116] Please refer to Figure 7 , Figure 7 It is a schematic structural diagram of an embodiment of the electronic device of the present application. The electronic device includes: a memory 60 and a processor 70 that are coupled to each other. Program instructions are stored in the memory 60, and the processor 70 is configured to execute the program instructions to implement the methods described in any of the above embodiments. Specifically, the electronic device includes, but is not limited to: desktop computers, laptop computers, tablet computers, servers, etc., which are not limited herein. In addition, the processor 70 may also be referred to as a CPU (Center Processing Unit, central processing unit). The processor 70 may be an integrated circuit chip with signal processing capabilities. The processor 70 may also be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. Additionally, the processor 70 may be implemented jointly by integrated circuit chips.

[0117] Please refer to Figure 8 , Figure 8 It is a schematic structural diagram of an embodiment of the computer-readable storage medium of the present application. Program instructions 90 that can be run by a processor are stored on the computer-readable storage medium 80, and when the program instructions 90 are executed by the processor, the methods described in any of the above embodiments are implemented.

[0118] In several embodiments provided by the present application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules or units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other may be through some interfaces, and the indirect couplings or communication connections of the devices or units may be in electrical, mechanical, or other forms.

[0119] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0120] In addition, in each embodiment of the present application, each functional unit can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0121] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods of each embodiment of the present application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0122] The above are only the embodiments of the present application, and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present application.< / useranswer> < / stdanswer> < / content> < / useranswer> < / stdanswer> < / content> < / useranswer> < / stdanswer>

Claims

1. A method for correcting answer content, characterized in that: include: Obtaining a target image corresponding to the current answer sheet, and determining at least one target partition corresponding to the target image using a recognition model; Based on the position information of the target partition in the target image, generate recognition prompt information of the target partition matching, input the recognition prompt information into the recognition large model, and generate the recognition result of the target image matching; Based on the recognition result, generating a target correction result that matches the answer content in the recognition result; Among them, the generating of a target correction result matching the answer content in the recognition result based on the recognition result includes: obtaining a reference answer matching the answer content, and obtaining an initial correction result matching the answer content based on the reference answer; obtaining a diversion judgment result based on the initial correction result; in response to the diversion judgment result determining that the initial correction result meets a preset condition, taking the initial correction result as the target correction result; in response to the diversion judgment result determining that the initial correction result does not meet the preset condition, generating correction prompt information based on the answer content and the reference answer, inputting the correction prompt information into the intelligent analysis model, and generating the target correction result matching the answer content.

2. The method according to claim 1, characterized in that The step of obtaining a target image corresponding to the current answer sheet and determining at least one target partition corresponding to the target image by using a recognition model includes: Capturing the current answer sheet to obtain the target image; Using the trained region detection module to obtain image coding features corresponding to the target image; The image coding features are input into the recognition model to generate at least one target partition corresponding to the target image; wherein the target partition is matched with the corresponding position information, and the position information includes the reference coordinates corresponding to the target partition.

3. The method according to claim 2, characterized in that The collecting the current answer sheet to obtain the target image includes: Acquire an initial image obtained by capturing the current answer sheet; wherein the initial image is matched with a corresponding initial length and initial width; A reference correction length matching the initial length and a reference correction width matching the initial width are obtained, and the initial image is resized based on the reference correction length and the reference correction width to obtain the adjusted target image.

4. The method according to claim 1, characterized in that The step of generating recognition prompt information of the target partition matching based on the position information of the target partition in the target image, inputting the recognition prompt information into the recognition large model, and generating the recognition result of the target image matching includes: Acquire a target recognition pattern that matches the target image; Generate the recognition prompt information based on the target recognition mode and the position information corresponding to the target partition; The recognition prompt information is input into the recognition large model to generate a structured recognition result matching the target image; wherein the recognition result includes recognition content matching at least one recognition category.

5. The method according to claim 1, characterized in that The step of generating a target correction result matching the answer content in the recognition result based on the recognition result includes: Obtaining a reference answer that matches the question content; Based on the answer content and the reference answer, correction prompt information is generated, and the correction prompt information is input into the intelligent analysis model to generate the target correction result matching the answer content.

6. The method according to claim 1, characterized in that The obtaining of the diversion judgment result based on the initial correction result includes: Acquire feature information matching the answer content; wherein the feature information includes at least one of the question stem information matching the answer content, the number of characters corresponding to the reference answer, and the language category; The initial correction result and the feature information are input into the trained diversion judgment model to obtain the diversion judgment result output by the diversion judgment model.

7. A system for correcting answer content, characterized in that: include: An acquisition module, used for acquiring a target image corresponding to the current answer sheet, and determining at least one target partition corresponding to the target image by using a recognition large model; A recognition module, used for generating recognition prompt information of the target partition matching based on the position information of the target partition in the target image, inputting the recognition prompt information into the recognition large model, and generating a recognition result of the target image matching; A processing module, configured to generate a target correction result matching the answer content in the recognition result based on the recognition result; Among them, the generating of a target correction result matching the answer content in the recognition result based on the recognition result includes: obtaining a reference answer matching the answer content, and obtaining an initial correction result matching the answer content based on the reference answer; obtaining a diversion judgment result based on the initial correction result; in response to the diversion judgment result determining that the initial correction result meets a preset condition, taking the initial correction result as the target correction result; in response to the diversion judgment result determining that the initial correction result does not meet the preset condition, generating correction prompt information based on the answer content and the reference answer, inputting the correction prompt information into the intelligent analysis model, and generating the target correction result matching the answer content.

8. An electronic device, characterized in that: include: A memory and a processor are coupled to each other, wherein the memory stores program instructions, and the processor is used to execute the program instructions to implement the method for correcting answer content as described in any one of claims 1-6.

9. A computer-readable storage medium having program instructions stored thereon, characterized in that: When the program instructions are executed by the processor, the method for correcting the answer content as described in any one of claims 1-6 is implemented.

Images