Answer extraction method and device

Abstract

The embodiment of the invention provides an answer extraction method and device.The answer extraction method comprises the steps that a homework map uploaded by a user is received, the homework map is uploaded after the user prints and uses a question sheet map, and the question sheet map is generated based on typesetting of multiple questions selected from a question bank; a matching graph corresponding to the homework graph is inquired based on the associated information of the homework graph, the matching graph is generated based on the question structure of each question in the question sheet graph, and different question areas in the matching graph have different display parameters; extracting region features of the job map and the matching map, and performing feature mapping on the job map and the matching map to obtain a region overlapping result of the job map and the matching map; and according to the region overlapping result, identifying a target answer in the answer region. The limitation of extracting the answers in a mode of presetting mark points is avoided, and the accuracy and efficiency of extracting the answers in the homework are higher by introducing the matching graph.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of image processing technology, and more particularly to a method and device for extracting an answer. Background Art

[0002] With the continuous advancement of technology, the efficiency of computer-assisted homework grading has been significantly improved compared to traditional manual grading methods. Therefore, in the field of education, automated homework grading technology has become increasingly important. However, in existing technical applications, students' homework answers are usually extracted through optical character recognition (OCR) technology. This method relies on specific marks in the questions to distinguish different questions, but these marks are easily disturbed by writing traces, thus affecting the accuracy of recognition. Therefore, it is urgent to provide a solution to the above technical problems. Summary of the invention

[0003] In view of this, an embodiment of this specification provides an answer extraction method. One or more embodiments of this specification also relate to an answer extraction device, a computing device, a computer-readable storage medium and a computer program to solve the technical defects existing in the prior art.

[0004] According to a first aspect of an embodiment of this specification, a method for extracting an answer is provided, comprising: Receive a homework diagram uploaded by a user, wherein the homework diagram is uploaded by the user after printing a question list diagram, and the question list diagram is generated based on the layout of multiple questions selected from a question bank; Based on the associated information of the homework graph, query the matching graph corresponding to the homework graph, wherein the matching graph is generated based on the question structure of each question in the question list graph, and different question areas in the matching graph have different display parameters; Extracting regional features of the operation graph and the matching graph, and performing feature mapping on the operation graph and the matching graph to obtain a regional overlap result of the operation graph and the matching graph; According to the region overlap result, a target answer in the answer region is identified.

[0005] Optionally, before receiving the job graph uploaded by the user, the step further includes: In response to the question setting request, selecting a plurality of questions from a question bank; According to the selection order of the multiple questions and preset layout information, the multiple questions are laid out to generate a question list diagram.

[0006] Optionally, after typeset the multiple topics, the method further includes: For any question, based on the question structure of the question, different display parameters are used to render different question areas of the question to generate a matching graph.

[0007] Optionally, based on the title structure of any of the titles, using different display parameters to render different title areas of any of the titles to generate a matching graph includes: Based on the title structure of any one of the titles, different first color parameters are used to render different title areas of any one of the titles to generate corresponding rectangular frames, and obtain a matching graph.

[0008] Optionally, also include: According to the topic types of the multiple topics, different second color parameters are used to perform background rendering on the ranges covered by the multiple topics respectively.

[0009] Optionally, after generating the matching graph, the step further includes: The question list graph and the matching graph are associated and stored, and the associated information of the question list graph is recorded, wherein the associated information of the question list graph corresponds to the associated information of the homework graph; The querying of a matching graph corresponding to the operation graph based on the association information of the operation graph includes: Based on the association information of the operation graph, a matching graph associated with the question sheet graph corresponding to the association information of the operation graph is queried from the stored matching graphs.

[0010] Optionally, the association information is specified by the user when uploading the homework diagram, or is carried in the homework diagram and the question list diagram in the form of a feature code.

[0011] Optionally, before extracting the regional features of the operation graph and the matching graph, the method further includes: Based on the associated information of the homework graph, query the question sheet graph corresponding to the homework graph; The extracting the regional features of the operation graph and the matching graph, and performing feature mapping on the operation graph and the matching graph to obtain the regional overlap result of the operation graph and the matching graph includes: Extracting a first regional feature of the homework graph, a second regional feature of the question list graph, and a third regional feature of the matching graph; Calculate a mapping matrix according to the first region feature and the second region feature; According to the mapping matrix and the third region feature, the matching graph is mapped to the operation graph to obtain a region overlap result between the operation graph and the matching graph.

[0012] Optionally, calculating a mapping matrix according to the first region feature and the second region feature includes: Extracting first target feature points from the first region feature and extracting second target feature points from the second region feature according to similarities between feature points in the first region feature and the second region feature; A mapping matrix is ​​calculated based on the first target feature points and the second target feature points.

[0013] Optionally, identifying a target answer in the answer area according to the area overlap result includes: Determining the answer area from the area overlap results based on display parameters corresponding to the answer area; A target answer is identified in the answer region.

[0014] According to a second aspect of an embodiment of this specification, there is provided an answer extraction device, comprising: A data receiving module is configured to receive an operation diagram uploaded by a user, wherein the operation diagram is uploaded after the user prints out a question list diagram and uses it, and the question list diagram is generated based on the layout of multiple questions selected from a question bank; A data query module is configured to query a matching graph corresponding to the homework graph based on the associated information of the homework graph, wherein the matching graph is generated based on the question structure of each question in the question list graph, and different question areas in the matching graph have different display parameters; A data mapping module is configured to extract regional features of the operation graph and the matching graph, and perform feature mapping on the operation graph and the matching graph to obtain a regional overlap result of the operation graph and the matching graph; The result identification module is configured to identify the target answer in the answer area according to the area overlap result.

[0015] According to a third aspect of an embodiment of this specification, a computing device is provided, including: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the above-mentioned answer extraction method are implemented.

[0016] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, and when the instructions are executed by a processor, the steps of the above-mentioned answer extraction method are implemented.

[0017] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the above-mentioned answer extraction method.

[0018] An embodiment of the present specification can generate a corresponding question list graph and matching graph based on the structure of the question. The matching graph can distinguish different question areas through different display parameters, thereby avoiding the need to use image recognition in the prior art to represent different question areas. By overlapping the matching graph with the work graph to segment the content in the work graph, the efficiency of extracting answers in the work graph is improved. In addition, the embodiment of the present specification does not need to pre-set special marks for each question, but distinguishes different question areas through different display parameters in the matching graph, so that it can adapt to more hierarchical question structures, does not occupy the question list graph layout to save space, and makes the answer extraction less susceptible to interference from the user's answer traces. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is a flow chart of an answer extraction method provided by an embodiment of this specification; Figure 2 is a schematic diagram of a question sheet diagram in an answer extraction method provided in one embodiment of the present specification; Figure 3 is a schematic diagram of a matching graph in an answer extraction method provided by an embodiment of this specification; Figure 4 It is a schematic diagram of overlapping of a work graph and a matching graph in an answer extraction method provided by an embodiment of the present specification; Figure 5 It is a processing flow chart of an answer extraction method provided by an embodiment of this specification and applied in an English homework correction scenario; Figure 6 is another processing flow chart of an answer extraction method provided by an embodiment of the present specification; Figure 7 is a schematic diagram of the structure of an answer extraction device provided by an embodiment of this specification; Figure 8 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0020] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0021] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0022] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0023] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0024] First, the terms involved in one or more embodiments of this specification are explained.

[0025] Affine Transformation: A linear mapping in geometry that preserves the collinearity between points and the distance ratio between parallel lines. This means that under an affine transformation, straight lines remain straight lines and parallel lines remain parallel.

[0026] Perspective Transformation: Also known as Projective Transformation, it is a nonlinear transformation that maps points in one plane to another plane. Unlike affine transformation, perspective transformation not only preserves the property of straight lines as straight lines, but also simulates the change of viewing angle in three-dimensional space, thus achieving a more realistic visual effect. It is widely used in computer vision, image processing and graphics, such as image correction, panorama stitching, virtual reality, etc.

[0027] HOG (Histogram of Oriented Gradients): is a computer vision technology used to describe the local shape and appearance characteristics of an image. The core idea of ​​HOG feature extraction is to characterize the edge information of the target by capturing the direction and amplitude of the change in pixel intensity in the image, thereby effectively describing the object.

[0028] In traditional school homework scenarios, paper is usually used as a carrier for classroom exercises and homework. In order to improve the efficiency of homework grading and summarize the students' answers, it is necessary to use automated equipment to detect and identify various information in paper homework. Existing technologies usually identify questions on paper based on OCR technology, and distinguish between question stem text and answer text, find the corresponding answer to the question in the question bank based on the question stem text, and then compare it with the answer text to obtain the grading result.

[0029] However, this method relies on special markers pre-set for the questions to distinguish different question areas, and the markers are easily disturbed. For example, problems may occur during the printing of the question sheet, resulting in interference and unclear marking of the markers. Or, if the traces of students’ answers or teachers’ annotations overlap with the markers, they will also interfere with the marker detection. Failure or errors in marker detection will affect the segmentation of the questions, resulting in low accuracy in answer recognition. In addition, in order to ensure the detection effect, space needs to be reserved for the markers, and markers need to be added for each question. Therefore, space will be wasted by markers before each question number.

[0030] To this end, in this specification, a method for extracting an answer is provided. This specification also relates to an answer extraction device, a computing device, and a computer-readable storage medium, which are described in detail one by one in the following embodiments.

[0031] See also Figure 1 , Figure 1 It is a flowchart of an answer extraction method provided by an embodiment of this specification, which specifically includes the following steps.

[0032] Step 102: receiving an assignment diagram uploaded by a user, wherein the assignment diagram is a printout of a question list diagram uploaded by the user, and the question list diagram is generated based on the layout of multiple questions selected from a question bank.

[0033] Specifically, the homework image refers to an image of the question sheet image printed and answered by the user. The image can be photographed or scanned by an image acquisition device and uploaded. The image acquisition device is, for example, a camera equipped in an electronic device such as a mobile phone or tablet. The user can be understood as a student answering the questions. The question sheet image is an image containing multiple questions.

[0034] Based on this, in order to automatically correct the user's homework, it is necessary to first obtain the homework picture uploaded by the user. The homework picture also corresponds to a question list picture, which is used to assist in extracting the answer from the homework picture.

[0035] Furthermore, in order to meet the user's various learning needs, the multiple questions in the question list diagram can be determined according to the user's independent selection. Before receiving the homework diagram uploaded by the user, it also includes: responding to the question group request, selecting multiple questions from the question bank; according to the selection order of the multiple questions and preset layout information, the multiple questions are typeset to generate a question list diagram.

[0036] Specifically, the question request is initiated by the user when he needs to practice questions. The user can be a teacher or a student. The question request can carry one or more of the subject information, question type, knowledge point type, and question quantity. A large number of different types of questions are pre-stored in the question bank. The preset layout information refers to the question arrangement rules set according to different teaching needs or user preferences. These rules may include the horizontal starting position of the question, the vertical starting position, the spacing between questions, the line spacing, the font size, the color, etc. The preset layout information can be set by the user after the question list is generated.

[0037] Based on this, the user initiates a question group request according to the needs, and in response to the question group request, multiple questions corresponding to the question group request are selected in the question bank. Since the order of selection of each question is different, this will also cause different questions to have different positions in the question list diagram. According to the selection order of the questions and the preset layout information, multiple questions are laid out to generate a question list diagram.

[0038] For example, see Figure 2 , Figure 2 It is a schematic diagram of a question list diagram in an answer extraction method provided by an embodiment of the present specification. The subject information carried by the user in the question setting request is English, and the knowledge point type is the past tense of vocabulary and the application of phrases. Based on this information, the user is shown questions that meet the requirements in the question bank, and the user's selection operation on the question is received. The corresponding question "Write the past tense of the verbs in the following phrases" is selected from the question bank, and the selection order of the question is recorded as 1. The preset typesetting information is "vertical starting position: 4th row, horizontal starting position: 1st character, font size: 4, line spacing: 1.5 times". The question selected by the user is typeset according to the selection order and the preset typesetting information, and the vertical starting position is updated according to the number of rows of the current question. For example, the current question occupies 5 rows in the typesetting, and the result obtained by adding the number of rows occupied by the question to the current vertical starting position is 9. The vertical starting position is updated to the 9th row, indicating that the next question is typeset from the 9th row.

[0039] In summary, the embodiments of this specification realize the rapid generation of personalized question list diagrams according to the specific needs of users. In this way, teachers can prepare teaching materials more efficiently and students can also get more targeted practice opportunities. By setting preset layout information, the layout of the question list diagram is more flexible and can adapt to different sizes of paper, reducing paper waste.

[0040] Furthermore, in order to more accurately extract the answers in the homework diagram without relying on markings, the embodiment of the present specification generates a matching diagram while generating a question list diagram. After the multiple questions are typeset, it also includes: for any question, based on the question structure of any question, different display parameters are used to render different question areas of any question to generate a matching diagram.

[0041] Specifically, the question structure refers to the structure contained in the question, answer space, illustrations, options, etc., or the question structure can also represent the hierarchy of the question, for example, a big question or a small question. Since the questions are stored in a structured manner in the question bank, the question structure can be directly obtained from the storage information of the question bank. The display parameters include color parameters, shape parameters, etc. The question area refers to the space occupied by different question structures after typeset.

[0042] Based on this, any question may contain multiple question structures. In order to extract the answer easily, it is necessary to accurately identify the different structures in the question. Therefore, different display parameters can be used to render different question areas. The question areas can be distinguished according to the question structure, and different question structures correspond to different question areas.

[0043] Furthermore, in order to be able to more clearly distinguish different question areas in a question, the question structure based on any one of the questions is used to render different question areas of any one of the questions using different display parameters to generate a matching graph, including: based on the question structure of any one of the questions, different first color parameters are used to render different question areas of any one of the questions to generate corresponding rectangular boxes to obtain a matching graph.

[0044] Specifically, the first color parameter refers to a color parameter used to distinguish different question structures. There may be multiple first color parameters, and the multiple first color parameters are different.

[0045] Based on this, since each question may contain multiple question structures. After the user selects a question from the question bank, a matching graph corresponding to the question graph can be generated at the same time. Since the questions are stored in a structured manner in the question bank, structuring means that the question stem, answer, analysis and other information in the question are stored in different fields in the form of key-value pairs. The sub-questions attached to the question are nested and stored in the sub-question field of the question structure in the same format as the question, such as: { "Question content": "Question text, in which the answer area $answer space 1$$answer space 2$ will be replaced by special marks, $latex formula$ and $table$ will be recorded in a special format, and the place where the picture needs to be inserted will be indicated by $picture name 1$"; "Question answer":["Answer to blank 1","Answer to blank 2"]; "Question Analysis": "Explanation and analysis of the answer"; "Sub-topic":[] } According to the storage information of the question in the question bank, it is possible to determine which question structures are included in the question, and then the different question areas corresponding to each question structure are represented by rectangular boxes of different colors.

[0046] Using the above example, see Figure 3 , Figure 3 is a schematic diagram of a matching graph in an answer extraction method provided by an embodiment of the present specification. For example, for the first question selected by the user, according to the storage information of the question in the question bank, it can be determined that the question stem of the question is "Write the past tense of the verbs in the following phrases." "(1) stand on my head$answer blank 1$""(2) hurt on my neck$Answer blank2$""(3) fall down$Answer blank3$""(4) go to hospital$Answer blank 4$”…, the question areas corresponding to different question structures are rendered in character order. The question area corresponding to the question stem “Write the past tense of the verbs in the following phrases.” (left diagonal fill area) can be rendered with a dark blue rectangle. The question stems corresponding to different types of questions can be distinguished by different colors, for example Figure 3 The first and fourth questions in the figure belong to different types of questions. The stem of the fourth question (the right diagonal filled area) is rendered with a red rectangle, and the question areas (diamond grid filled areas) corresponding to multiple answer spaces such as $answer space 1$, $answer space 2$, and $answer space 3$ can be represented by green rectangles. Similarly, the question areas of other questions can be rendered in this way, and finally a matching image corresponding to the question sheet image is obtained.

[0047] In summary, the embodiments of the present specification can determine the question structure of a question based on the stored information of the question without scanning the question, thus avoiding the problem of inaccurate question recognition by traditional image recognition methods and having higher robustness. Different colors are used to render different question areas according to different question structures, and intuitive visual feedback is used to distinguish the question structure. The embodiments of the present specification not only improve the efficiency of answer extraction, but also do not require the setting of explicit marking points and do not occupy the question sheet layout. No matter how many levels of structure there are, they can be represented by adding new rectangular boxes. The technical implementation is simple, the effect is stable, and the user-side display information is not changed, so the user is not aware of it, thereby improving the user experience.

[0048] Furthermore, in order to distinguish different types of questions, the answer extraction method also includes: according to the question types of the multiple questions, using different second color parameters to perform background rendering on the ranges covered by the multiple questions respectively.

[0049] Specifically, the question types include multiple choice questions, fill-in-the-blank questions, true-or-false questions, and short answer questions, etc. The second color parameter refers to a color parameter used to distinguish different question types.

[0050] Based on this, since the types of multiple questions selected by the user may be different, the questions can be stored with corresponding type tags in the question bank. After obtaining the question selected by the user, the question type of the question can be determined according to the type tag in the storage information of the question. Each question type uses different colors to render the background of the range it covers according to its question type.

[0051] Using the above example, see Figure 3 ,Among the questions selected by the user, the first question is a fill-in-the-blank question, and its background (area filled with dots) is rendered in light blue, and the last question is a comprehensive question, and its background (area filled with vertical lines) is rendered in white.

[0052] In summary, the embodiments of the present specification distinguish the types of questions and render the coverage of different types of questions through different color parameters, which can more accurately distinguish the questions and facilitate homework grading. In addition, this method has no restrictions on the distinction of question types. No matter how many types of questions there are, they can be distinguished by simply changing the corresponding rendering colors without taking up additional space.

[0053] Furthermore, in order to ensure that the position of the questions in the matching graph is accurate and corresponds to the position of the questions in the question list graph, during the rendering of the matching graph, it is also necessary to render it in combination with the preset layout information. The preset layout information includes the horizontal starting position, vertical starting position, spacing between questions, font size, color, etc.

[0054] Specifically, multiple questions are rendered in sequence according to the order in which the questions are selected. For the first question, the horizontal indentation pixels when each question is rendered are determined according to the horizontal starting position of the question; the vertical indentation pixels when the question is rendered are determined according to the vertical starting position. Specifically, the vertical axis where the pixel corresponding to the vertical starting position is located is used as the vertical indentation pixel. The number of rows occupied by the question can be determined according to the line spacing and font size. The pixel range covered by the question can be determined according to the horizontal indentation pixels, the vertical indentation pixels and the number of rows occupied by the question. The pixel coverage range includes the horizontal pixel coverage range and the vertical pixel coverage range. Then, the pixel range covered by the question is rendered as a background, and the vertical indentation pixels are added to the vertical pixel coverage range, and the vertical indentation pixels are updated according to the result of the addition. Subsequent questions can be rendered in the same way.

[0055] Continuing with the above example, the horizontal starting position of the question determines that the corresponding horizontal indentation pixels are 1000 pixels, and the vertical starting position determines that the vertical indentation pixels are 2000 pixels. According to the line spacing and font size, the horizontal coverage pixel range of the question is 6000 pixels, and the vertical pixel coverage range is 2000 pixels. Based on the above parameters, the background of the area covered by the question is rendered. After the rendering is completed, the vertical indentation pixels need to be updated to 8000 pixels so that the next question can be rendered according to the updated vertical indentation pixels.

[0056] In summary, the embodiments of the present specification make the position of the question in the matching graph more accurate by accurately calculating and updating the pixel coverage range and indentation pixels during the question rendering process, and the matching graph uses the same parameters as the question list graph generation process, so that the question positions represented by different colors in the finally generated matching graph can completely overlap with the corresponding questions in the question list graph, thereby improving the accuracy of the final extracted answer.

[0057] Furthermore, in order to facilitate the subsequent matching of the homework graph with the matching graph, after the matching graph is generated, it also includes: associating and storing the question list graph and the matching graph, and recording the association information of the question list graph, wherein the association information of the question list graph corresponds to the association information of the homework graph.

[0058] Specifically, since the homework image is the image after printing out the question sheet image and answering the questions, the content of the homework image and the associated question sheet image is the same. The associated information of the question sheet image may include the matching image identifier of the matching image associated with the question sheet image, or may also be the homework name, homework identifier, etc.

[0059] Based on this, after the question list graph and the matching graph are generated, the question list graph and the matching graph are associated and stored, and the associated information of the question list graph is recorded. The associated information of the question list graph can be carried in the question list graph.

[0060] In summary, the embodiments of the present specification can store the question list graph and the matching graph in association with each other, so as to facilitate the subsequent positioning of the corresponding question list graph and the matching graph according to the association information of the homework graph, and ensure that in the subsequent answer extraction process, the corresponding matching graph can be found quickly and accurately through the association information in the homework graph, thereby realizing automated and high-precision answer extraction.

[0061] Step 104: Based on the associated information of the homework graph, query the matching graph corresponding to the homework graph, wherein the matching graph is generated based on the question structure of each question in the question list graph, and different question areas in the matching graph have different display parameters.

[0062] Specifically, the association information is used to characterize the association relationship between the homework graph, the matching graph, and the question list graph. Based on this, in order to facilitate the query of the matching graph associated with the homework graph, the question list graph carries the corresponding related information, so the homework graph also carries the corresponding related information. After the user uploads the homework graph, the related information in the homework graph is identified, and the corresponding matching graph is queried based on the identified related information.

[0063] Furthermore, since the contents of the homework image and the question list image are the same, but the homework image may be displaced or tilted during the printing process, resulting in the inability to overlap the homework image and the matching image, it is also necessary to find the corresponding question list image when searching for the matching image, so as to facilitate the subsequent adjustment of the matching image through the question list image so that the adjusted matching image can overlap with the homework image. Therefore, querying the matching image corresponding to the homework image based on the association information of the homework image includes: querying the matching image associated with the question list image corresponding to the association information of the homework image from the stored matching image based on the association information of the homework image.

[0064] Specifically, during the implementation process, the associated information in the homework diagram is first identified. Then, the associated information is compared with the question map associated information stored in the database to find the question map that matches the homework diagram associated information. Finally, based on the matched question map, the matching map associated with it is further queried, thereby achieving effective correspondence between the homework diagram and the matching map. It should be noted that the search order of the question map and the matching map can be either sequential or simultaneous. When querying sequentially, this manual does not limit the search order of the question map and the matching map.

[0065] In summary, the embodiments of this specification use the pre-set association information to query the question list image and the matching image, so that the query results are more accurate. Even if the job image is displaced or tilted during the printing process, the question list image can be adjusted with the assistance of the question list image to ensure the overlap between the matching image and the job image, thereby improving the accuracy and efficiency of answer extraction. The matching image facilitates the subsequent segmentation of the question structure in the job image, thereby extracting the answer corresponding to the question.

[0066] Furthermore, the association information may be in various forms. The association information is specified by the user when uploading the homework diagram, or is carried in the homework diagram and the question list diagram in the form of a feature code.

[0067] Specifically, specifying the associated information when the user uploads the homework map can be understood as the user clearly writing the homework name or homework identifier at a specific location in the homework map after completing the questions in the homework map.

[0068] Continuing with the above example, before uploading the assignment diagram, the user can write "Assignment ID: E-001" in the assignment diagram. In this case, after receiving the assignment diagram, it is necessary to identify the assignment ID E-001 carried in the assignment diagram, and then find the question list diagram with the same ID from the stored question list diagram. Since the question list diagram is stored in association with the matching diagram, the corresponding matching diagram can be obtained after finding the question list diagram.

[0069] Alternatively, the associated information can also be carried in the work map in the form of a feature code. The feature code can be in the form of a QR code, barcode, etc. After the user uploads the work map, the feature code in the work map is identified to obtain the identifier of the question list map associated with the work map or the identifier of the matching map.

[0070] Continuing with the above example, the work drawing will carry a QR code when printed. The QR code carries the associated information of the question list drawing and the matching drawing associated with the work drawing. After identifying the QR code, the information obtained is E-001, and then the question list drawing and the matching drawing with the same identification information and the representation information are selected.

[0071] In summary, the embodiments of this specification ensure the accurate correspondence between the homework map, the question list map, and the matching map by flexibly using various forms of associated information. Whether it is the homework identification manually specified by the user or the associated information automatically identified by the feature code, the corresponding question list map and matching map can be quickly located, thereby greatly improving the efficiency and accuracy of answer extraction.

[0072] Step 106: extracting regional features of the operation graph and the matching graph, and performing feature mapping on the operation graph and the matching graph to obtain a regional overlap result of the operation graph and the matching graph.

[0073] Specifically, the regional feature refers to an image feature corresponding to the region where the job content is located in the job image or the matching image.

[0074] Based on this, in order to accurately extract the answers in the homework image, it is necessary to distinguish the answers from the questions. The matching image can already represent the different question areas in the question sheet image corresponding to the homework image through different display effects. Therefore, feature mapping is performed on the homework image and the matching image, that is, the homework image and the matching image are overlapped, and the regional features in the homework image are matched with the regional features in the matching image. Finally, the regional overlap result of the homework image and the matching image can be obtained.

[0075] Furthermore, since the homework image may be displaced, tilted, etc. during the printing process, the homework image and the matching image cannot completely overlap, which will lead to inaccurate extraction of the answers in the homework image. Therefore, before extracting the regional features of the homework image and the matching image, it also includes: based on the association information of the homework image, querying the question list image corresponding to the homework image.

[0076] Correspondingly, after searching for the corresponding question list diagram, a mapping matrix can be calculated according to the mapping relationship between the question list diagram and the homework diagram, and then the matching diagram can be corrected according to the calculated mapping matrix. Only after the correction can the overlap between the homework diagram and the matching diagram be higher, thereby improving the accuracy of answer extraction in the homework diagram. The extracting of regional features of the homework diagram and the matching diagram, and performing feature mapping on the homework diagram and the matching diagram to obtain the regional overlap result of the homework diagram and the matching diagram includes: extracting the first regional feature of the homework diagram, the second regional feature of the question list diagram, and the third regional feature of the matching diagram; calculating the mapping matrix according to the first regional feature and the second regional feature; mapping the matching diagram to the homework diagram according to the mapping matrix and the third regional feature to obtain the regional overlap result of the homework diagram and the matching diagram.

[0077] Specifically, the first regional feature is the regional feature of different regions in the questions in the homework map, and the feature can also reflect the relative position relationship between the region where the question is located and the entire homework map. The second regional feature is the regional feature of different regions in the questions in the question list map, and the feature can also reflect the relative position relationship between the region where the question is located and the entire question list map. The third regional feature is the regional feature of different regions in the questions in the matching map, and the feature can also reflect the relative position relationship between the region where the question is located and the entire matching map. The mapping matrix is ​​a matrix that can accurately describe the spatial transformation relationship between the homework map and the question list map. The mapping matrix can include basic transformation parameters such as translation, rotation, and scaling, as well as correction factors for nonlinear deformations to ensure that high-precision feature alignment can still be achieved in complex situations. The mapping matrix can be an affine transformation matrix or a perspective transformation matrix.

[0078] Based on this, after finding the question list image and matching image corresponding to the homework image, in order to improve the matching degree between the homework image and the question list image, it is necessary to first extract the first area features of the homework image, extract the second area features of the question list image, and extract the third area features of the matching image. The mapping matrix is ​​calculated based on the correspondence and position differences between the above-mentioned first area features and the second area features. Since the area where the questions in the question list image are located is consistent with the area where the questions in the matching image are located, by applying the mapping matrix and the third area features corresponding to the matching image, the question area in the matching image can be effectively and accurately aligned with the corresponding area in the homework image, thereby obtaining the regional overlapping effect of the homework image and the matching image.

[0079] Using the above example, see Figure 4 , Figure 4 It is a schematic diagram of the overlap of an assignment image and a matching image in an answer extraction method provided by an embodiment of the present specification, wherein the first region feature in the assignment image and the second region feature in the question sheet image are extracted by the HOG feature extraction method, a mapping matrix is ​​calculated by the first region feature and the second region feature, and the matching image is adjusted by the mapping matrix, so that the matching image overlaps with the assignment image, and after the overlap, the overall coverage of the first question in the assignment image coincides with the light blue background in the matching image, the question area (left diagonal filled area) where the stem of the first question in the assignment image "write the past tense of the verbs in the following phrases" is located coincides with the dark blue rectangular box in the matching image, the question area (diamond grid filled area) where the answer space for the first question in the assignment image is located coincides with the green rectangular box in the matching image, and the background area (dot filled area) of the first question in the assignment image coincides with the light blue rectangular box in the matching image.

[0080] In summary, the embodiments of this specification significantly improve the overlap between the operation image and the matching image through accurate mapping matrix calculation and feature alignment. In practical applications, this method can effectively deal with various deformation problems caused by the operation image during scanning and transmission, making the answer extraction more accurate.

[0081] Furthermore, in order to calculate the mapping matrix more accurately and efficiently, it is necessary to extract feature points with high similarity from the homework image and the question list image respectively, and calculate the mapping matrix through these feature points. Therefore, the calculation of the mapping matrix based on the first area feature and the second area feature includes: extracting the first target feature point from the first area feature and the second target feature point from the second area feature according to the similarity of the feature points in the first area feature and the second area feature; and calculating the mapping matrix based on the first target feature point and the second target feature point.

[0082] Specifically, the first target feature point is a key feature point identified in the first area feature corresponding to the work diagram, and the second feature point is a key feature point identified in the second area feature corresponding to the matching diagram. The first target feature point and the corresponding second target feature point are feature points with high similarity. The number of the first target feature points and the second target feature points can be multiple.

[0083] Based on this, the feature points in the first area feature and the second area feature are corresponding. If the mapping matrix is ​​calculated based on the correspondence between all the feature points in the first area feature and the second area feature, the amount of calculation will be very large, resulting in low calculation efficiency. Therefore, in order to improve the calculation efficiency without reducing the accuracy of the mapping matrix, the target feature points with higher similarity can be extracted based on the similarity between the feature points in the first area feature and the second area feature. Specifically, the first feature point in the first area feature can be extracted, and the second feature point corresponding to the first feature point in the second area feature can be extracted, and the similarity between the first feature point and the second feature point can be calculated. When the similarity is greater than the preset similarity threshold, the first feature point is used as the first target feature point, and the second feature point is used as the second target feature point. Similarly, other first feature points in the first area feature and other second feature points in the second area feature can be judged in the above manner whether they can be used as the first target feature point and the second target feature point.

[0084] In another embodiment of the present specification, the similarities may be sorted in descending order, and a preset number of first feature points with a higher order are used as first target feature points, and a preset number of second feature points with a higher order are used as second target feature points.

[0085] Afterwards, the first target feature point and the second target feature point may be input into a first preset function to calculate a mapping matrix.

[0086] Furthermore, the mapping matrix and the third region feature are input into a second preset function to transform the matching graph.

[0087] Continuing with the above example, there are modules in the Python open source library OpenCV that support affine transformation (cv2.getAffineTransform and cv2.warpAffine) and perspective transformation (cv2.getPerspectiveTransform and cv2.warpPerspective).

[0088] Take perspective transformation as an example: Assume that there are first feature points A1, B1, C1... in the homework image, and second feature points A2, B2, C2... in the question image, among which A1 and A2 are similar feature points with a similarity of 98%, B1 and B2 are similar feature points with a similarity of 96%, C1 and C2 are similar feature points with a similarity of 95%..., take the first three first feature points A1, B1, C1 as the first target feature points, and take the second three second feature points A2, B2, C2 as the second target feature points, cv2.getPerspectiveTransform function accepts n coordinate points in the original image (that is, all the first target feature points) and the corresponding n coordinates in the question image (that is, all the second target feature points), processes these feature points and returns a perspective transformation matrix M. cv2.warpPerspective transforms the matching image into a form similar to the homework image through the matrix M obtained above, so that the overlap between the homework image and the matching image is higher, thereby improving the accuracy of answer extraction. The same is true for affine transformation.

[0089] In summary, the embodiments of this specification significantly improve the efficiency and accuracy of generating the mapping matrix by accurately selecting and calculating the target feature points. This method not only reduces the consumption of computing resources, but also greatly improves the accuracy of answer extraction. It avoids the problem of inaccurate answer extraction caused by displacement or deformation factors during the printing process.

[0090] Step 108: Identify the target answer in the answer area according to the area overlap result.

[0091] Specifically, the area overlapping results can show the segmentation effect of different question structures and different question types in the questions. According to the segmentation effect, the answer area in the homework image can be determined, and the corresponding answers can be extracted from the answer area to realize the correction of the questions in the homework image.

[0092] The step of identifying a target answer in the answer area according to the area overlap result includes: determining the answer area from the area overlap result based on display parameters corresponding to the answer area; and identifying the target answer in the answer area.

[0093] Specifically, the target answer refers to the answer filled in by the user in the job graph.

[0094] Based on this, since the various question structures in the question have different colors when rendered, the answer area corresponds to specific display parameters. By analyzing the display parameters such as pixel distribution or color in the area overlap results, the system can accurately locate the boundary of the answer area. After that, it identifies the specific target answer in the determined answer area and compares it with the standard answer, thereby completing the automatic correction of the questions in the homework image.

[0095] Continuing with the above example, after the homework graph overlaps with the matching graph, the first three questions are fill-in-the-blank questions, which belong to the same type of questions and overlap with the light blue background area of ​​the matching graph. The stem of the first question "1. Write the past tense of the verbs in the following phrases" overlaps with the dark blue rectangular area in the matching graph, and the answer area overlaps with the green rectangular area in the matching graph. Since the display parameter of the answer area is a green rectangle, the area where the green rectangle is located can be found from the overlapping results, and the answer can be identified. According to the different display parameters and the corresponding question structure in the matching graph, the various questions in the homework graph and the question structure in the questions are segmented. After segmentation, the answers filled in by the user in the first three fill-in-the-blank questions are extracted, such as: "stood", "hurt", "fall", "went"..., and the answers "F", "D", "B", "C" are extracted from the fourth comprehensive question.

[0096] In summary, the embodiments of this specification not only improve the efficiency of grading, but also ensure the accuracy of the grading results.

[0097] The following combination Figure 5 , taking the application of the answer extraction method provided in this specification in the English homework correction scenario as an example, the answer extraction method is further explained. Among them, Figure 5 This is a processing flow chart of an answer extraction method provided by an embodiment of this specification and applied in an English homework grading scenario, which specifically includes the following steps.

[0098] Step 502: In response to the question setting request, multiple questions are selected from the question bank.

[0099] Step 504: Layout the multiple questions according to the selection order of the multiple questions and preset layout information to generate a question list diagram.

[0100] Step 506: for any question, based on the question structure of the any question, different first color parameters are used to render different question areas of the any question to generate corresponding rectangular frames, and obtain a matching graph.

[0101] Step 508: According to the topic types of the multiple topics, different second color parameters are used to perform background rendering on the ranges covered by the multiple topics respectively.

[0102] Step 510: The question list graph and the matching graph are stored in association, and the association information of the question list graph is recorded, wherein the association information of the question list graph corresponds to the association information of the homework graph.

[0103] Step 512: Based on the association information of the operation graph, query the matching graph associated with the question sheet graph corresponding to the association information of the operation graph from the stored matching graphs.

[0104] Step 514: Based on the associated information of the homework graph, query the question sheet graph corresponding to the homework graph.

[0105] Step 516: Extract the first regional features of the homework graph, the second regional features of the question list graph, and the third regional features of the matching graph.

[0106] Step 518: extracting first target feature points from the first region feature and extracting second target feature points from the second region feature according to the similarity between the feature points in the first region feature and the second region feature.

[0107] Step 520: Calculate a mapping matrix based on the first target feature points and the second target feature points.

[0108] Step 522: Map the matching graph to the operation graph according to the mapping matrix and the third region feature, and obtain a region overlap result between the operation graph and the matching graph.

[0109] Step 524: Determine the answer area from the area overlap results based on the display parameters corresponding to the answer area.

[0110] Step 526: Identify a target answer in the answer region.

[0111] In summary, the questions in the embodiments of this specification are stored in a structured manner in the question bank. Based on the structure of the questions, corresponding question list graphs and matching graphs can be generated. The matching graphs can display different question areas through different display parameters, thereby avoiding the need to represent different question areas through image recognition in the prior art. The content in the work graph is segmented by overlapping the matching graph with the work graph, thereby improving the efficiency of extracting answers in the work graph. Furthermore, the matching graph is adjusted by the mapping matrix calculated by the question list graph and the work graph, so that the overlap between the adjusted matching graph and the work graph is higher, thereby avoiding the problem that the work graph cannot overlap with the matching graph due to displacement, tilt, etc. during the answering process, thereby further improving the accuracy of answer extraction. In addition, the embodiments of this specification do not need to pre-set special marks for each question, but distinguish different question types or different question structures through different colors, so that it can adapt to more question types and question structures, save space for the work graph, and make the answer extraction less susceptible to interference from the user's answer traces.

[0112] See also Figure 6 , Figure 6 This is another processing flow chart of an answer extraction method provided by an embodiment of the present specification, including steps 602 to 616.

[0113] Step 602: User group question list.

[0114] Users can select multiple topics from the topics and combine them according to their needs.

[0115] Step 604: Generate a question list graph and a matching graph.

[0116] A question list graph and a corresponding matching graph are generated according to the multiple questions selected by the user. The specific generation process is shown in the aforementioned embodiment and will not be repeated in this embodiment.

[0117] Step 606: The user prints the question sheet image for use and uploads the assignment image.

[0118] The user prints the question sheet and answers the questions therein. After completing the answers, the worksheet is captured by an image acquisition device and uploaded.

[0119] Step 608: Obtain the question list graph and matching graph.

[0120] Query the question list graph and matching graph corresponding to the homework graph uploaded by the user. For the specific query process, please refer to the above embodiment, which will not be repeated in this embodiment.

[0121] Step 610: Calculate the mapping matrix between the homework graph and the question list graph.

[0122] The mapping matrix is ​​calculated based on the first area features corresponding to the homework diagram and the second area features corresponding to the question list diagram. The specific calculation process is shown in the above embodiment and will not be repeated in this embodiment.

[0123] Step 612: Overlay the matching graph and the operation graph according to the mapping matrix.

[0124] The matching graph is adjusted according to the mapping matrix, and the adjusted matching graph is superimposed on the operation graph.

[0125] Step 614: Segment the title frame based on the matching graph.

[0126] After the matching graph and the operation graph are superimposed, the rectangular boxes of different colors in the matching graph will correspond to the corresponding question structures in the operation graph, so that different types of questions and different types of question structures can be segmented.

[0127] Step 616: Use the segmented question information and answer information to make corrections.

[0128] After segmentation, the question information (stem information) and answer information (answers) in the homework graph can be extracted, so that the questions can be corrected based on the extracted information.

[0129] In summary, the questions in the embodiments of this specification are stored in a structured form in the question bank. Based on the structure of the question, the corresponding question list graph and matching graph can be automatically generated. The matching graph displays different question areas through different display parameters, which effectively avoids the limitation of the prior art that image recognition must be relied on to distinguish different question areas. By overlapping the matching graph with the operation graph, the content in the operation graph can be effectively segmented, thereby improving the efficiency of extracting answers. Furthermore, the matching graph is adjusted by the mapping matrix calculated by the question list graph and the operation graph, so that the adjusted matching graph has a higher degree of overlap with the operation graph, which solves the problems of displacement and tilt of the operation graph that may occur during the answering process, and further improves the accuracy of answer extraction. In addition, the embodiments of this specification do not need to pre-set special marks for each question, but distinguish different question types or structures by different colors, so as to adapt to a wider range of question types and structures, save space for the operation graph, and reduce interference from user answer traces.

[0130] Corresponding to the above method embodiment, this specification also provides an answer extraction device embodiment, Figure 7 FIG. 2 shows a schematic diagram of a structure of an answer extraction device provided by an embodiment of the present specification. Figure 7 As shown, the device comprises: The data receiving module 702 is configured to receive a worksheet uploaded by a user, wherein the worksheet is uploaded by the user after printing out a question sheet, and the question sheet is generated based on the layout of multiple questions selected from a question bank; The data query module 704 is configured to query the matching graph corresponding to the homework graph based on the association information of the homework graph, wherein the matching graph is generated based on the question structure of each question in the question list graph, and different question areas in the matching graph have different display parameters; A data mapping module 706 is configured to extract regional features of the operation graph and the matching graph, and perform feature mapping on the operation graph and the matching graph to obtain a regional overlap result of the operation graph and the matching graph; The result identification module 708 is configured to identify the target answer in the answer area according to the area overlap result.

[0131] The answer extraction device further comprises: The question list diagram generation module is configured to select multiple questions from the question bank in response to a question grouping request; layout the multiple questions according to the selection order of the multiple questions and preset layout information to generate a question list diagram.

[0132] The matching graph generation module is configured to render different question areas of any question using different display parameters based on the question structure of any question, so as to generate a matching graph.

[0133] The matching graph generation module is further configured to render different question areas of any question using different first color parameters based on the question structure of any question to generate corresponding rectangular frames to obtain a matching graph.

[0134] The matching graph generation module is further configured to use different second color parameters to perform background rendering on the ranges covered by the multiple questions respectively according to the question types of the multiple questions.

[0135] The answer extraction device further comprises: A storage module is configured to associate and store the question list graph with the matching graph, and record association information of the question list graph, wherein the association information of the question list graph corresponds to the association information of the homework graph; The data query module 704 is further configured to query the matching graph associated with the question sheet graph corresponding to the associated information of the homework graph from the stored matching graphs based on the associated information of the homework graph.

[0136] The association information is specified by the user when uploading the homework diagram, or is carried in the homework diagram and the question list diagram in the form of a feature code.

[0137] The data query module 704 is further configured to query the question list graph corresponding to the homework graph based on the associated information of the homework graph; The data mapping module 706 is further configured to extract the first regional features of the homework graph, the second regional features of the question list graph, and the third regional features of the matching graph; calculate a mapping matrix based on the first regional features and the second regional features; map the matching graph to the homework graph based on the mapping matrix and the third regional features to obtain a regional overlap result between the homework graph and the matching graph.

[0138] The data mapping module 706 is further configured to extract first target feature points from the first area feature and second target feature points from the second area feature according to the similarity of feature points in the first area feature and the second area feature; and calculate a mapping matrix based on the first target feature points and the second target feature points.

[0139] The result identification module 708 is further configured to determine the answer area from the area overlap results based on the display parameters corresponding to the answer area; and identify the target answer in the answer area.

[0140] The above is a schematic scheme of an answer extraction device of this embodiment. It should be noted that the technical scheme of the answer extraction device and the technical scheme of the answer extraction method described above belong to the same concept, and the details not described in detail in the technical scheme of the answer extraction device can be referred to the description of the technical scheme of the answer extraction method described above.

[0141] Figure 8 The block diagram of a computing device 800 according to an embodiment of the present specification is shown. The components of the computing device 800 include but are not limited to a memory 810 and a processor 820. The processor 820 is connected to the memory 810 via a bus 830, and a database 850 is used to store data.

[0142] The computing device 800 also includes an access device 840 that enables the computing device 800 to communicate via one or more networks 860. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 840 may include one or more of any type of network interface (e.g., a network interface card (NIC)) of wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a world-wide interoperability for microwave access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, and a near field communication (NFC).

[0143] In one embodiment of the present specification, the above components of the computing device 800 and Figure 8 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 8 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0144] The computing device 800 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 800 may also be a mobile or stationary server.

[0145] The processor 820 is used to execute the following computer executable instructions, which, when executed by the processor, implement the steps of the above-mentioned answer extraction method.

[0146] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above answer extraction method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above answer extraction method.

[0147] An embodiment of the present specification also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the above-mentioned answer extraction method.

[0148] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the answer extraction method described above are of the same concept, and the details not described in detail in the technical scheme of the storage medium can be found in the description of the technical scheme of the answer extraction method described above.

[0149] An embodiment of the present specification also provides a computer program, wherein when the computer program is executed in a computer, the computer is caused to execute the steps of the above-mentioned answer extraction method.

[0150] The above is a schematic scheme of a computer program of this embodiment. It should be noted that the technical scheme of the computer program and the technical scheme of the above answer extraction method belong to the same concept, and the details not described in detail in the technical scheme of the computer program can be found in the description of the technical scheme of the above answer extraction method.

[0151] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0152] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0153] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0154] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0155] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. A method for extracting an answer, characterized in that: include: Receive a homework diagram uploaded by a user, wherein the homework diagram is uploaded by the user after printing a question list diagram, and the question list diagram is generated based on the layout of multiple questions selected from a question bank; Based on the associated information of the homework graph, query the matching graph corresponding to the homework graph, wherein the matching graph is generated based on the question structure of each question in the question list graph, and different question areas in the matching graph have different display parameters; Extracting regional features of the operation graph and the matching graph, and performing feature mapping on the operation graph and the matching graph to obtain a regional overlap result of the operation graph and the matching graph; According to the region overlap result, a target answer in the answer region is identified.

2. The answer extraction method according to claim 1, characterized in that: Before receiving the job graph uploaded by the user, the method further includes: In response to the question setting request, multiple questions are selected from the question bank; According to the selection order of the multiple questions and preset layout information, the multiple questions are laid out to generate a question list diagram.

3. The answer extraction method according to claim 2, characterized in that: After typeset the multiple topics, the method further includes: For any question, based on the question structure of the question, different display parameters are used to render different question areas of the question to generate a matching graph.

4. The answer extraction method according to claim 3, characterized in that: The method of rendering different topic areas of any topic based on the topic structure of any topic using different display parameters to generate a matching graph includes: Based on the title structure of any one of the titles, different first color parameters are used to render different title areas of any one of the titles to generate corresponding rectangular frames, and obtain a matching graph.

5. The answer extraction method according to claim 4, characterized in that: Also includes: According to the topic types of the multiple topics, different second color parameters are used to perform background rendering on the ranges covered by the multiple topics respectively.

6. The answer extraction method according to claim 3, characterized in that: After generating the matching graph, the method further includes: The question list graph and the matching graph are associated and stored, and the associated information of the question list graph is recorded, wherein the associated information of the question list graph corresponds to the associated information of the homework graph; The querying of a matching graph corresponding to the operation graph based on the association information of the operation graph includes: Based on the association information of the operation graph, a matching graph associated with the question sheet graph corresponding to the association information of the operation graph is queried from the stored matching graphs.

7. The answer extraction method according to claim 1 or 6, characterized in that: The association information is specified by the user when uploading the homework diagram, or is carried in the homework diagram and the question list diagram in the form of a feature code.

8. The answer extraction method according to claim 1, characterized in that: Before extracting the regional features of the operation graph and the matching graph, the method further includes: Based on the associated information of the homework graph, query the question sheet graph corresponding to the homework graph; The extracting the regional features of the operation graph and the matching graph, and performing feature mapping on the operation graph and the matching graph to obtain the regional overlap result of the operation graph and the matching graph includes: Extracting a first regional feature of the homework graph, a second regional feature of the question list graph, and a third regional feature of the matching graph; Calculate a mapping matrix according to the first region feature and the second region feature; According to the mapping matrix and the third region feature, the matching graph is mapped to the operation graph to obtain a region overlap result between the operation graph and the matching graph.

9. The answer extraction method according to claim 8, characterized in that: The calculating a mapping matrix according to the first regional feature and the second regional feature comprises: Extracting first target feature points from the first region feature and extracting second target feature points from the second region feature according to similarities between feature points in the first region feature and the second region feature; A mapping matrix is ​​calculated based on the first target feature points and the second target feature points.

10. The answer extraction method according to any one of claims 1-6, 8 and 9, characterized in that: The step of identifying a target answer in the answer area according to the area overlap result comprises: Determining the answer area from the area overlap results based on display parameters corresponding to the answer area; A target answer is identified in the answer region.

11. An answer extraction device, characterized in that: include: A data receiving module is configured to receive an operation diagram uploaded by a user, wherein the operation diagram is uploaded after the user prints out a question list diagram and uses it, and the question list diagram is generated based on the layout of multiple questions selected from a question bank; A data query module is configured to query a matching graph corresponding to the homework graph based on the associated information of the homework graph, wherein the matching graph is generated based on the question structure of each question in the question list graph, and different question areas in the matching graph have different display parameters; A data mapping module is configured to extract regional features of the operation graph and the matching graph, and perform feature mapping on the operation graph and the matching graph to obtain a regional overlap result of the operation graph and the matching graph; The result identification module is configured to identify the target answer in the answer area according to the area overlap result.

12. A computing device, characterized in that: include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the answer extraction method described in any one of claims 1 to 10 are implemented.

13. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, can implement the steps of the answer extraction method described in any one of claims 1 to 10.

14. A computer program product, characterized in that It comprises computer instructions, which, when executed by a processor, implement the steps of the answer extraction method described in any one of claims 1 to 10.

Images