Methods, devices, equipment and media for grading questions

Abstract

This disclosure relates to a method, apparatus, device, and medium for grading exam questions, comprising: acquiring an image of a question page; identifying printed text in the question page image; retrieving a standard page image corresponding to the same question page in a preset first database based on the printed text; retrieving a target reference answer corresponding to a target question in the question page image based on the standard page image in a preset second database, wherein the second database includes reference answers marked with keywords; and grading the answer content corresponding to the target question based on the target reference answer. This disclosure can improve the accuracy of question grading.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of image processing, and particularly relates to a question correction method and device, equipment and medium. BACKGROUND

[0002] Photograph correction is a common correction method in education and teaching. At present, photograph correction can be accurately and efficiently performed on general questions with fixed answers, but some questions have the same answers but various answering methods, such as common application questions, which can adopt various solving processes to obtain the same answers. It is very complex to correct such questions, and the current photograph correction method has low accuracy. SUMMARY

[0003] In order to solve the above technical problems or at least partially solve the above technical problems, the present disclosure provides a question correction method, device, equipment and medium.

[0004] According to an aspect of the present disclosure, a question correction method is provided, comprising:

[0005] obtaining a question page image;

[0006] recognizing printed text in the question page image;

[0007] retrieving a standard page image corresponding to the same question page as the question page image in a preset first database according to the printed text;

[0008] retrieving a target reference answer corresponding to the question title in the question page image in a preset second database based on the standard page image, wherein the second database includes reference answers marked with keywords;

[0009] correcting the answer content corresponding to the question title according to the target reference answer.

[0010] According to another aspect of the present disclosure, a question correction device is provided,

[0011] an image acquisition module configured to obtain a question page image;

[0012] a text recognition module configured to recognize printed text in the question page image;

[0013] an image retrieval module configured to retrieve a standard page image corresponding to the same question page as the question page image in a preset first database according to the printed text;

[0014] The answer retrieval module is used to retrieve, based on the standard page image, a target reference answer corresponding to the target question in a preset second database for the target question in the question page image. The second database includes reference answers marked with keywords.

[0015] The grading module is used to grade the answers to the target questions based on the target reference answers.

[0016] According to another aspect of this disclosure, an electronic device is provided, the electronic device comprising:

[0017] processor;

[0018] Memory used to store the processor's executable instructions;

[0019] The processor is configured to read the executable instructions from the memory and execute the instructions to implement the above-mentioned question correction method.

[0020] According to another aspect of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, which, when executed on a terminal device, cause the terminal device to implement a question grading method.

[0021] The technical solution provided in this disclosure has the following advantages compared with the prior art:

[0022] The question grading method, apparatus, device, and medium provided in this disclosure include: acquiring a question page image; identifying printed text in the question page image; retrieving a standard page image corresponding to the same question page in a preset first database based on the printed text; retrieving a target reference answer corresponding to the target question in a preset second database based on the standard page image for the target question in the question page image, wherein the second database includes reference answers marked with keywords; and grading the answer content corresponding to the target question based on the target reference answer. This disclosure can improve the accuracy of question grading. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0024] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart of a question grading method provided in this embodiment of the present disclosure;

[0026] Figure 2 A schematic diagram of the title page image and standard page image provided for embodiments of this disclosure;

[0027] Figure 3 This is a schematic diagram illustrating the correction of answer content for different answer types provided in the embodiments of this disclosure;

[0028] Figure 4 This is a schematic diagram of the structure of the judgment model provided in the embodiments of this disclosure;

[0029] Figure 5 This is a schematic diagram of the structure of the question correction device provided in the embodiments of this disclosure;

[0030] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0031] To better understand the above-described objects, features, and advantages of this disclosure, embodiments of the disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of the disclosure are shown in the drawings, it should be understood that the disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0032] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0033] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below. It should be noted that the concepts of "first", "second", etc., used in this disclosure are only used to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0034] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0035] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0036] For word problems and similar questions with identical answers but varying responses, current photo-based grading methods are cumbersome and have low accuracy. Therefore, to support photo-based grading of word problems and similar question types and improve grading accuracy, this disclosure provides a method, apparatus, device, and medium for grading questions. For ease of understanding, the embodiments of this disclosure are described in detail below.

[0037] Figure 1 A flowchart of a question grading method is provided, which can be executed by an electronic device or a server. Both the electronic device and the server are question grading terminals in this embodiment. The electronic device may include devices with communication capabilities such as smartphones, tablets, desktop computers, and laptops. The server may be a cloud server or a server cluster, or other devices with storage and computing capabilities.

[0038] like Figure 1 As shown, the grading method for this question may include the following steps:

[0039] S102. Obtain the image of the question page.

[0040] In this embodiment, the question page image can be obtained through image selection, image capture, or image upload. The question page image includes at least one question, such as fill-in-the-blank, true / false, drawing, and application questions. Each question includes printed question content and handwritten answer content.

[0041] In a scenario using homework as an example, electronic devices such as mobile phones and tablets can be used to photograph the pages of students' completed homework assignments. The photographed images can then undergo image processing such as correction, noise reduction, and light compensation to obtain high-quality images of the homework pages. These images can be saved on the electronic device and used for subsequent homework correction, or they can be uploaded to a server for correction.

[0042] This embodiment utilizes images of the question page for subsequent test grading, which can shorten grading time, improve grading efficiency, and avoid the influence of human factors on the grading results, thereby improving grading accuracy. Furthermore, in practical applications, taking photos or selecting question page images from an image library is very convenient, without time or location restrictions, allowing grading to be done anytime, anywhere, thus improving grading convenience.

[0043] S104. Identify printed text in the image of the question page.

[0044] In one embodiment, the fonts in the title page image can be classified using a pre-trained font classification model to obtain the printed font category; the printed area images in the title page image that belong to the printed font category can be determined; and then the text recognition of each printed area image can be performed using a preset text recognition model to obtain the printed text.

[0045] S106. Based on the printed text, retrieve the standard page image corresponding to the same title page image from the preset first database.

[0046] In this embodiment, during retrieval, a standard page image corresponding to the same title page as the title page image can be retrieved from the first database storing standard page images based on the printed text in the title page image and / or the layout information between the titles; the standard page image is an electronic image of the same title page as the title page image.

[0047] like Figure 2 For example, the image on the left shows the actual answer to the question page, while the image on the right shows the electronic version of the same question page. The standard page image and the question page image contain the same question and have the same page layout.

[0048] Because standard page images greatly narrow down the search scope of questions, it becomes much easier to retrieve subsequent reference answers based on standard page images.

[0049] S108. For the target question in the question page image, retrieve the target reference answer corresponding to the target question from the preset second database based on the standard page image. The second database includes reference answers marked with keywords.

[0050] In this embodiment, the target question can be any question in the question page image. It can be understood that there is a one-to-one correspondence between the standard page image and the questions in the question page image. Therefore, firstly, the current standard question corresponding to the target question in the question page image is determined in the standard page image. Then, the current reference answer corresponding to the current standard question is obtained from the second database storing reference answers, and this current reference answer is determined as the target reference answer corresponding to the target question.

[0051] The aforementioned second database stores reference answers marked with keywords, and establishes a correlation between these reference answers and the standard questions. Based on this correlation, reference answers associated with the current standard question can be quickly retrieved from the second database. The keywords marked in the reference answers are generally numbers, subjects, etc., found within the answer. For example, in the reference answer "Xiaoming and Xiaofang each have 3 apples," the keyword could be the number "3." Keywords not only express the key and unchanging information in the reference answer but also, through word segmentation, do not restrict the order or format of the answer. Therefore, keywords can serve as effective information for question grading. When grading question types with fixed answers but flexible solutions (such as word problems), only a comparison of some words in the solution with the keywords is needed for grading. This grading method reduces the difficulty of grading and improves accuracy.

[0052] Existing technologies typically retrieve the current standard question corresponding to the target question directly from the question database, one question at a time, and then use the reference answer of the current standard question as the target reference answer. However, this method of finding questions individually becomes very slow when there are many questions; moreover, for questions with little text, such as drawing questions, it is difficult to find the accurate current standard question. Compared to the method of finding questions by question, the solution provided in steps S106 and S108 of this embodiment is to first start from the overall question page image to retrieve the standard page image. Since the overall page has many features suitable for matching (such as text information, layout information, and similar graphic features), it is possible to quickly and accurately retrieve the matching standard page image; moreover, for drawing questions with little text, using images to retrieve images can reduce the difficulty of retrieval and improve the retrieval accuracy. After retrieving the standard page image, the search scope of the questions is greatly narrowed, thereby determining the current standard question of the target question based on the standard page image, and assigning the reference answer of the current standard question to the target question, which can obtain the target reference answer quickly and accurately. Therefore, the above method of first retrieving the image and then determining the reference answer can improve the efficiency and accuracy of answer retrieval.

[0053] S110. Based on the target reference answer, correct the answers to the target questions.

[0054] In this embodiment, the question area of ​​the target question in the question page image can be detected first, and then the key text content in the question area can be detected. The key text content includes characters such as "answer" or "solution", underscores, and parentheses. The area where the key text content in the question area is located is determined as the answer area. The answer content in the answer area is then identified.

[0055] In a possible implementation, the solution content can be compared with the target reference answer. If the answers match, the solution content is considered correct. If the answers do not match, it does not necessarily mean that the solution content is incorrect; rather, the solution content may be an alternative to the target reference answer. Since the keywords in the answer are fixed regardless of the solution method, this implementation can further compare the word segmentation in the solution content with the keywords marked in the target reference answer. If the words match, the solution content is considered correct; if the words do not match, the solution content is considered incorrect.

[0056] The question grading method provided in this embodiment acquires a question page image; identifies printed text in the question page image; retrieves a standard page image corresponding to the same question page in a preset first database based on the printed text; for a target question in the question page image, retrieves a target reference answer corresponding to the target question in a preset second database based on the standard page image, wherein the second database includes reference answers marked with keywords; and grades the answer content corresponding to the target question based on the target reference answer. In this technical solution, firstly, acquiring the question page image and grading the questions based on it can greatly shorten grading time and improve grading efficiency; secondly, retrieving a standard page image based on the printed text in the question page image, and then retrieving the target reference answer based on the standard page image, can improve retrieval efficiency and accuracy; nextly, grading the answer content based on the target reference answer marked with keywords, wherein the keywords can express the key and unchanging answer information in the target reference answer, and are not limited to a fixed answer format in the form of word segmentation; thus, using the target reference answer and its keywords for grading can reduce grading difficulty and improve grading accuracy.

[0057] Regarding step S108 above, this embodiment provides a specific process for retrieving the target reference answer corresponding to the target question from a preset second database based on a standard page image, including the following:

[0058] The process involves detecting the question region of a target question within a question page image. One approach is to detect the question number and text line information in the image; determine the boundary coordinates of each text line based on these information; determine the area of ​​each question based on these boundary coordinates; and finally, determine the question region of the current target question based on the area of ​​each question. Alternatively, the question page image can be input into a pre-trained segmentation model, which outputs a bounding box enclosing the question region, thus obtaining the target question region. These are merely examples of question region detection; other detection methods can be used in practical applications.

[0059] Identify the standard questions corresponding to the question area in the standard page image; it can be understood that there is a one-to-one correspondence between the questions in the standard page image and the questions in the question page image. Therefore, identify the standard questions corresponding to the question area in the question page image in the standard page image.

[0060] Obtain a second database associated with the standard page image; retrieve the reference answers for the standard questions from the second database; and determine the retrieved reference answers as the target reference answers for the target questions.

[0061] In this embodiment, a corresponding reference answer can be generated for each standard question in the standard page image, and the keywords of the reference answer can be marked. The reference answers marked with keywords are stored in a second database, and an association is established between the standard page image and the second database, as well as an association is established between each standard question and its reference answer. In this case, the reference answer for the standard question can be retrieved from the second database, and this reference answer can be determined as the target reference answer for the target question.

[0062] The reference answers in this embodiment can be obtained in the following ways. A reference answer for each question can be provided in a standard page image, such as a teacher's guide; alternatively, the answer area can be left blank, similar to the question page. For standard page images providing reference answers, the text-formatted reference answer for each question can be directly extracted and associated with the corresponding standard question. For standard page images with blank answer areas, refer to... Figure 2 The reference answers are manually written in the answer field of each question. To facilitate the retrieval of the reference answers, the complete reference answers are also manually written for fill-in-the-blank questions. Then, the text-formatted reference answers for each question are extracted and associated with the corresponding standard questions.

[0063] In this embodiment, keywords in the reference answer can be manually marked; alternatively, keywords in the reference answer can be extracted using keyword extraction algorithms, and the reference answer can be marked with these keywords. These keyword extraction algorithms include TF-IDF (Term Frequency-Inverse Document Frequency), TextRank, and RAKE (Rapid Automatic Keyword Extraction).

[0064] Based on the above method, when determining the standard question corresponding to the target question, it is possible to obtain the reference answer marked with keywords corresponding to the standard question, and then determine the reference answer as the target reference answer corresponding to the target question.

[0065] In this embodiment, based on the target reference answers marked with keywords, the answers corresponding to the target questions are corrected. The implementation process of this step can be referred to... Figure 3 As shown.

[0066] This embodiment includes: obtaining the answer type of the solution content in the target question; and grading the solution content according to the answer type.

[0067] In this context, the standard question or target reference answer corresponding to the target question can be pre-marked with the answer type, which is also the answer type of the solution content. Alternatively, the answer type can be obtained by classifying the solution content.

[0068] The above-mentioned answer types can generally be divided into: mixed types containing text and numbers, graphical types, purely numerical types, and purely text types. Different grading methods can be used for different answer types, which will be introduced below.

[0069] In this embodiment, in response to determining that the answer type is a mixed answer type containing text and numbers, the answer content is graded according to the target reference answer and its marked keywords by using a preset judgment model.

[0070] The answer type includes a mixture of text and numbers, such as the answer "each segment is 4 meters long". For mixed-type answers, this embodiment can use a judging model to grade the answers based on the target reference answer and its marked keywords.

[0071] In one example, the judging model is, for instance, the BERT (Bidirectional Encoder Representations from Transformers) model. BERT is a pre-trained language model based on the Transformer architecture, whose main goal is to learn general language representations through pre-training, allowing for fine-tuning across various natural language processing tasks.

[0072] BERT's pre-training process consists of two stages: Masked Language Model (MLM) and Next Sentence Prediction (NSP). In the MLM stage, BERT randomly replaces some words in the input sentence with "[MASK]" tags and then trains the model to predict these replaced words. In the NSP stage, BERT takes two sentences as input and trains the model to determine whether the two sentences are consecutive.

[0073] Through the two pre-training stages described above, BERT can learn contextual information within sentences, thereby better understanding their meaning. In the fine-tuning stage, BERT can be used for various natural language processing tasks, such as text classification, question answering systems, and named entity recognition. BERT achieves excellent performance in natural language processing tasks.

[0074] like Figure 4 As shown, the judging model in this embodiment may include an embedding module, an encoding module, and a decoding module. Based on this, the process of grading the answer content according to the target reference answer and its marked keywords through the preset judging model can be referred to as follows.

[0075] By using the embedding module, the answer content is mapped to an answer matrix, the target reference answer is mapped to a reference answer matrix, and the keywords of the target reference answer are mapped to a word matrix.

[0076] The encoding module determines the first semantic relevance between the solution matrix and the reference answer matrix, and the second semantic relevance between the solution matrix and the word matrix.

[0077] The decoding module identifies the semantic relevance between the solution content and the target reference answer based on the first relevance, and identifies whether the solution content contains keywords of the target reference answer based on the second relevance. Then, the solution content is corrected based on the semantic relevance and keywords.

[0078] According to this embodiment, the semantic relevance between the semantics of the answer content and the semantics of the target reference answer is identified based on the first relevance, and it is determined that the semantic relevance is higher than a preset relevance threshold. If the semantic relevance is not higher than the preset relevance threshold, it indicates that the semantics expressed by the answer content and the target reference answer are different, and the answer content can be determined to be incorrect. If the semantic relevance is higher than the preset relevance threshold, it indicates that the semantics expressed by the answer content and the target reference answer are the same, and a second relevance can be combined to improve the accuracy of grading. Therefore, the second relevance is used to identify whether the answer content contains keywords of the target reference answer. If the answer content contains keywords, the answer content is determined to be correct.

[0079] To avoid misjudgment, this embodiment can also generate a review prompt message when the semantic relevance is not higher than the relevance threshold and / or the answer content contains keywords. This review prompt message is used to prompt teachers, parents, and other users to manually judge whether the answer content is correct. Subsequently, this embodiment can receive user feedback on the judgment results of the answer content and mark the judgment results on the answer content.

[0080] This embodiment uses a judgment model and grades the answers based on the target reference answer and its keywords, which can improve the accuracy of grading and increase the support for complex question types such as word problems, thereby grading students' questions more efficiently.

[0081] In this embodiment, in response to determining that the answer type is a drawing type, the image similarity between the image in the target reference answer and the image in the answer content is determined; and the answer content is corrected based on the image similarity.

[0082] In this example, the answer type is a drawing type, such as drawing an equilateral triangle. For drawing-type answers, this embodiment can grade the question based on the similarity between images. Specifically, the first image in the answer and the second image in the target reference answer are extracted. For easier comparison, the first and second images can be scaled down to two images with similar sizes. The first and second images are processed into 256-dimensional matrices using the ResNet18 network model, and the Euclidean distance between the two matrices is calculated. The Euclidean distance is used to represent the image similarity between the first and second images. The smaller the Euclidean distance, the higher the image similarity. If the Euclidean distance is less than a preset distance threshold, that is, the image similarity is higher than a preset similarity value, the answer is determined to be correct; conversely, if the image similarity is not higher than the preset similarity value, the answer is determined to be incorrect.

[0083] In this embodiment, in response to determining that the answer type is purely numeric, the text matching degree between the target reference answer and the answer content is determined; and the answer content is corrected based on the text matching degree.

[0084] In this embodiment, in response to determining that the answer type is plain text, the answer content is identified; when the answer content is identified as a word, the answer content is corrected based on the semantic similarity between the target reference answer and the answer content; when the answer content is identified as a sentence, the answer content is corrected based on the target reference answer and its marked keywords.

[0085] This embodiment can identify information such as text length, semantics, part-of-speech tagging, and named entities in the answer content, and determine whether the answer content is a word or a sentence based on this information. If the answer content is identified as a word, then the corresponding target reference answer is also a word. Based on this, the answer content can be corrected according to the semantic similarity between the target reference answer and the answer content. It can be understood that a high semantic similarity between the target reference answer and the answer content indicates that the answer content and the target reference answer may be synonyms, such as "same" and "equal." In this case, the answer content can be determined to be correct.

[0086] If the answer is identified as a sentence, it can be reviewed against the mixed-type answer containing text and numbers, and corrected based on the target reference answer and its tagged keywords. For example, it can check if the semantic similarity between the answer and the target reference answer is higher than a preset relevance threshold; if it is, it can check if the word segments in the answer are semantically identical to the keywords in the target reference answer; if the word semantics are identical, the answer is considered correct. Conversely, if at least one of the above is incorrect, the answer is considered incorrect.

[0087] The above embodiments employ different grading methods for different types of answers, which can increase the support for grading more questions and improve the accuracy of question grading.

[0088] After grading each question according to the above embodiments, this embodiment can also calculate the total score of all questions on the entire question page image based on the preset score of each question. Furthermore, it may also include receiving user-submitted comments and suggestions. All the information, including the question page image, question grading results, total score, and comments, can be saved on an electronic device or uploaded to the cloud for easy viewing and tracking by the user.

[0089] In summary, the question grading method provided in this disclosure, which utilizes question page images for grading, can shorten grading time and improve grading efficiency, accuracy, and convenience. When retrieving standard page images based on printed text in the question page images, the overall page has many matching features, thus enabling quick and accurate retrieval of matching standard page images. Furthermore, for drawing questions with minimal text, using images to retrieve images reduces retrieval difficulty and improves accuracy. After retrieving the standard page images, the search scope for questions is significantly narrowed, thereby improving retrieval efficiency and accuracy by using standard page images to find target reference answers. When grading target questions, the answers are graded based on the target reference answers marked with keywords. Since the keywords in the target reference answers can express the key and unchanging information in the answer, and are not limited to a fixed solution format in the form of word segmentation, when grading question types with fixed answers but flexible solution methods (such as word problems), only a portion of the words in the solution content needs to be compared with the keywords to make a grade. This grading method can reduce the difficulty of grading and improve the accuracy of grading.

[0090] This disclosure also provides an apparatus for implementing the above-described question grading method, which is described below in conjunction with... Figure 5The following explanation is provided. In this embodiment, the question-correction device can be an electronic device or a server. The electronic device can include devices with communication capabilities such as mobile phones, tablets, desktop computers, and laptops. The server can be a cloud server or server cluster, or other devices with storage and computing capabilities.

[0091] Figure 5 A schematic diagram of the structure of a question correction device provided in an embodiment of this disclosure is shown.

[0092] like Figure 5 As shown, the question grading device 200 may include:

[0093] Image acquisition module 210 is used to acquire images of the question page;

[0094] Text recognition module 220 is used to recognize printed text in the title page image;

[0095] Image retrieval module 230 is used to retrieve, based on the printed text, a standard page image corresponding to the same title page image in a preset first database;

[0096] The answer retrieval module 240 is used to retrieve a target reference answer corresponding to the target question in the question page image based on the standard page image in a preset second database, wherein the second database includes reference answers marked with keywords;

[0097] The grading module 250 is used to grade the answers to the target questions based on the target reference answers.

[0098] In one embodiment, the answer retrieval module 240 is further configured to:

[0099] Detect the question region of the target question in the question page image;

[0100] Determine the standard question corresponding to the question area in the standard page image;

[0101] Obtain the second database associated with the standard page image;

[0102] Retrieve the reference answers to the standard questions from the second database;

[0103] The retrieved reference answers are determined as the target reference answers for the target question.

[0104] In one embodiment, the correction module 250 is further configured to:

[0105] Obtain the answer type of the solution content in the target question;

[0106] The answers are graded according to the given answer type.

[0107] In one embodiment, the correction module 250 is further configured to:

[0108] In response to determining that the answer type is a drawing type, the image similarity between the image in the target reference answer and the image in the answer content is determined;

[0109] The answer is corrected based on the image similarity.

[0110] The correction module 250 is further used for:

[0111] In response to determining that the answer type is a purely numeric type, the text matching degree between the target reference answer and the answer content is determined;

[0112] The answer content is corrected based on the text matching degree.

[0113] In one embodiment, the correction module 250 is further configured to:

[0114] In response to determining that the answer type is a mixed answer type containing text and numbers, the answer content is graded according to the target reference answer and its marked keywords using a preset judgment model.

[0115] In one embodiment, the judging model includes an embedding module, an encoding module, and a decoding module, and the grading module 250 is further configured to:

[0116] Through the embedding module, the answer content is mapped to an answer matrix, the target reference answer is mapped to a reference answer matrix, and the keywords of the target reference answer are mapped to a word matrix;

[0117] The encoding module determines the first semantic correlation between the solution matrix and the reference answer matrix, and the second semantic correlation between the solution matrix and the word matrix.

[0118] The decoding module identifies the semantic correlation between the semantics of the answer content and the semantics of the target reference answer based on the first correlation, and identifies whether the answer content contains the keywords of the target reference answer based on the second correlation.

[0119] The answer content is corrected based on the semantic relevance and the keywords.

[0120] In one embodiment, the correction module 250 is further configured to:

[0121] In response to determining that the response type is plain text, the content of the response is identified;

[0122] When the answer content is identified as a word, the answer content is corrected based on the semantic similarity between the target reference answer and the answer content.

[0123] When the answer is identified as a sentence, the answer is corrected based on the target reference answer and its marked keywords.

[0124] The device provided in this embodiment has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0125] Exemplary embodiments of this disclosure also provide an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the electronic device to perform a method according to an embodiment of this disclosure.

[0126] Exemplary embodiments of this disclosure also provide a computer program product, including a computer program, wherein, when executed by a processor of a computer, the computer program is used to cause the computer to perform a method according to an embodiment of this disclosure.

[0127] refer to Figure 6 The present invention describes a structural block diagram of an electronic device 300 that can serve as a server or client of the present disclosure, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0128] like Figure 6As shown, the electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 302 or a computer program loaded from a storage unit 308 into a random access memory (RAM) 303. The RAM 303 may also store various programs and data required for the operation of the device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0129] Multiple components in electronic device 300 are connected to I / O interface 305, including: input unit 306, output unit 307, storage unit 308, and communication unit 309. Input unit 306 can be any type of device capable of inputting information to electronic device 300. Input unit 306 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 307 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 308 may include, but is not limited to, disk and optical disk. Communication unit 309 allows electronic device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0130] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above. For example, in some embodiments, the question grading method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. In some embodiments, the computing unit 301 can be configured to perform the question grading method by any other suitable means (e.g., by means of firmware).

[0131] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0132] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0133] As used in this disclosure, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0134] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0135] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0136] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

[0137] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A question grading method, comprising: obtaining a question page image; recognizing printed text in the question page image; retrieving, according to the printed text, a standard page image corresponding to a same question page as the question page image in a preset first database; wherein the standard page image corresponding to the same question page as the question page image is retrieved in the first database storing standard page images according to layout information between the printed text and / or questions in the question page image; the standard page image corresponds to a question in the question page image one by one; retrieving, based on the standard page image, a target reference answer corresponding to a question title in the question page image in a preset second database, wherein the second database includes reference answers marked with keywords; the keywords are used to express key and fixed answer information in the reference answer; grading answer content corresponding to the question title according to the target reference answer; wherein word segmentation in the answer content is compared with the keywords marked in the target reference answer.

2. The method of claim 1, wherein, The retrieving, based on the standard page image, a target reference answer corresponding to a question title in the question page image in a preset second database, comprises: detecting a question area of the question title in the question page image; determining a standard question corresponding to the question area in the standard page image; obtaining a second database associated with the standard page image; retrieving a reference answer of the standard question in the second database; determining the retrieved reference answer as the target reference answer of the question title.

3. The method of claim 1 or 2, wherein, The grading answer content corresponding to the question title according to the target reference answer, comprises: obtaining an answer type of the answer content in the question title; grading the answer content according to the answer type.

4. The method of claim 3, wherein, The grading the answer content according to the answer type, comprises: in response to determining that the answer type is a drawing type, determining an image similarity between an image in the target reference answer and an image in the answer content; grading the answer content according to the image similarity, wherein the grading the answer content according to the answer type further comprises: in response to determining that the answer type is a pure number type, determining a text matching degree between the target reference answer and the answer content; grading the answer content according to the text matching degree.

5. The method of claim 3, wherein, The grading the answer content according to the answer type, comprises: in response to determining that the answer type is a mixed answer type containing words and numbers, grading the answer content according to the target reference answer and the keywords marked therein by a preset judging model.

6. The method of claim 5, wherein, The judging model comprises an embedding module, an encoding module and a decoding module. The grading the answer content according to the target reference answer and the keywords marked therein by a preset judging model, comprises: The embedding module maps the answer content as an answer matrix, maps the target reference answer as a reference answer matrix, and maps the keywords of the target reference answer as a keyword matrix; The encoding module determines a first correlation between the answer matrix and the reference answer matrix in a semantic level, and a second correlation between the answer matrix and the keyword matrix in a semantic level; The decoding module identifies a semantic correlation between the semantics of the answer content and the semantics of the target reference answer according to the first correlation, and identifies whether the target reference answer keywords are contained in the answer content according to the second correlation; The answer content is corrected according to the semantic correlation and the keywords.

7. The method of claim 3, wherein, The correcting the answer content according to the answer type comprises: In response to determining that the answer type is a pure text type, the answer content is identified; When the answer content is identified as a word, the answer content is corrected according to the semantic similarity between the target reference answer and the answer content; When the answer content is identified as a sentence, the answer content is corrected according to the target reference answer and the labeled keywords thereof.

8. A test grading apparatus, characterized by comprising: It comprises: An image acquisition module is configured to acquire a question page image; A text recognition module is configured to recognize printed text in the question page image; An image retrieval module is configured to retrieve a standard page image corresponding to the same question page as the question page image in a preset first database according to the printed text in the question page image; wherein the standard page image corresponding to the same question page as the question page image is retrieved in the first database storing the standard page image according to the layout information between the printed text in the question page image and / or the question; the standard page image corresponds to the question in the question page image one by one; An answer retrieval module is configured to retrieve a target reference answer corresponding to a question title in the question page image based on the standard page image in a preset second database; wherein the second database includes reference answers labeled with keywords; the keywords are used to express key and fixed answer information in the reference answers; A correction module is configured to correct the answer content corresponding to the question title according to the target reference answer; wherein the keywords in the answer content are compared with the keywords labeled in the target reference answer.

9. An electronic device, comprising: The electronic device comprises: A processor; A memory for storing executable instructions of the processor; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium having stored thereon computer instructions, wherein, When the computer instructions are run on a terminal device, the terminal device implements the method of any one of claims 1-7.

Images