Bulk filling question automatic marking method and system, storage medium and equipment

Abstract

The invention discloses an automatic gap filling question reviewing method and system, a storage medium and equipment, and the method comprises the steps: obtaining the original input information of an interface where a to-be-reviewed test question is located, carrying out the question scene classification of the to-be-reviewed test question, and at least determining the grade and subject of the to-be-reviewed test question; identifying an answer area corresponding to the to-be-read test question and an answer area corresponding to the to-be-read test question; the answer content in the answer area is extracted, the answer content in the answer area is extracted, and the answer content is classified; comparing the answer content with the answer content according to a reviewing rule corresponding to the category to which the answer content belongs, and outputting an initial reviewing result of the test question to be reviewed; and evaluating the initial review result by adopting a preset answer review model, and outputting a target review result. According to the invention, the problems of low marking efficiency and marking precision due to lack of cross-disciplinary and cross-grade overall planning and marking capability and mechanization of marking thinking in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, system, storage medium and device for automatically marking fill-in-the-blank questions. Background Art

[0002] There are a large number of automatic grading scenarios in homework and exams. Among them, the grading of fill-in-the-blank questions has not yet been realized on a large scale because it is impossible to achieve mechanical comparison with the answers like multiple-choice questions. It still needs to be done manually by teachers, which consumes a lot of educational resources.

[0003] According to public information, in the existing fill-in-the-blank question grading technology, there are relatively mature technical ideas for the formal classification of fill-in-the-blank questions. For example, questions can be divided into pure text type, pure numerical type, expression type, and other types. Different types of answer content and answer data can be classified from the perspective of grading, such as unique answer type (the text content must be exactly the same), formal equivalence type (the expression is equivalent after moving terms and simplification), semantic identity type (whether the semantic level is consistent), etc.

[0004] However, existing automatic marking solutions do not take into account scenario differences. In real automatic marking, there may only be two inputs: the student's answer content and the answer content, without the question stem content. For some questions with equivalent expressions or similar semantics, the judgments given are often different depending on the grade or subject to which the specific question belongs. In addition, existing automatic marking technologies often focus on processing one or several categories, and few can handle various forms of content judgment at the same time, and the ability to judge semantic similarity is insufficient.

[0005] Therefore, the existing automatic marking technology for fill-in-the-blank questions lacks the ability to conduct overall marking across disciplines and grades, and the marking ideas are relatively mechanical, resulting in low marking efficiency and accuracy. Summary of the invention

[0006] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method, system, storage medium and device for automatic marking of fill-in-the-blank questions, aiming to solve the problem that the automatic marking technology of fill-in-the-blank questions in the prior art lacks the ability to conduct overall planning and marking across disciplines and grades, and the marking ideas are relatively mechanical, resulting in low marking efficiency and accuracy.

[0007] A first aspect of the present invention is to provide a method for automatically marking fill-in-the-blank questions, the method comprising: Acquire original input information of the interface where the test questions to be reviewed are located, classify the test questions to be reviewed according to the original input information, and at least determine the grade and subject to which the test questions to be reviewed belong; Identify the answer area corresponding to the test question to be marked, and the answer area corresponding to the test question to be marked; Extracting answer content in the answer area according to the answer area, and extracting answer content in the answer area according to the answer area, and classifying the answer content; According to the review rules corresponding to the category to which the answer content belongs, the answer content is compared with the answer content, and an initial review result of the test question to be reviewed is output; Using a preset answer review model to evaluate the initial review result and output a target review result; The categories of the answer content include pure numerical value, character symbol, expression, professional terminology and text description; Among them, the step of comparing the answer content with the answer content according to the review rules corresponding to the category to which the answer content belongs, and outputting the initial review result of the test question to be reviewed includes: Comparing the answer content with the answer content according to the evaluation criteria corresponding to the category to which the answer content belongs; If the answer content is exactly the same as the answer content, then the answer content is judged to be correct; If the answer content is not completely the same as the answer content, the answer content is converted to obtain equivalent content, and the equivalent content is compared with the answer content to determine whether they are the same; According to the judgment result, the initial marking result corresponding to the test question to be marked is output.

[0008] The step of using a preset answer review model to evaluate the initial review result and outputting a target review result includes: Importing the original input information into a preset answer review model, so that the answer review model performs semantic understanding based on the original input information, and screens out review criteria corresponding to the original input information in a database; According to the review standard, the initial review result and / or the answer content corresponding to the professional terminology category and the text description category are reviewed, and the target review result is output.

[0009] According to one aspect of the above technical solution, the steps of obtaining original input information of the interface where the test questions to be reviewed are located, classifying the test questions to be reviewed according to the original input information, and at least determining the grade and subject to which the test questions to be reviewed belong include: Identify at least one interface where a test question to be reviewed is located, and obtain original input information corresponding to the test question to be reviewed in the target interface; According to the original input information, the test questions to be marked are classified by question scenarios to at least determine the grade and subject to which the test questions to be marked belong.

[0010] According to one aspect of the above technical solution, the steps of extracting answer content in the answer area according to the answer area, extracting answer content in the answer area according to the answer area, and classifying the answer content include: According to the answer area, determining whether the answer area is empty, and if not, extracting the answer content in the answer area; Extracting answer content in the answer area according to the answer area; The answer content is classified into forms according to a preset regular expression.

[0011] According to one aspect of the above technical solution, when the category of the answer content is a professional terminology category or a text description category, the step of classifying the answer content further includes: A pre-trained deep neural network is used to classify the answer content to determine the chapter features to which the answer content belongs.

[0012] According to one aspect of the above technical solution, the answer review model is an LLM model.

[0013] The second aspect of the present invention is to provide a fill-in-the-blank question automatic marking system, which is applied to the method described in the above technical solution, and the system comprises: A first classification module is used to obtain original input information of the interface where the test questions to be reviewed are located, and classify the test questions to be reviewed according to the original input information, and at least determine the grade and subject to which the test questions to be reviewed belong; An area identification module, used to identify the answer area corresponding to the test question to be marked, and the answer area corresponding to the test question to be marked; A second classification module is used to extract the answer content in the answer area according to the answer area, and to extract the answer content in the answer area according to the answer area, and classify the answer content; A result output module, used for comparing the answer content with the answer content according to the review rules corresponding to the category to which the answer content belongs, and outputting the initial review result of the test question to be reviewed; The result evaluation module is used to evaluate the initial review result by using a preset answer review model and output a target review result.

[0014] A third aspect of the present invention is to provide a readable storage medium having a computer program stored thereon, which implements the method described in the above technical solution when executed by a processor.

[0015] The fourth aspect of the present invention is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the method described in the above technical solution when executing the computer program.

[0016] Compared with the prior art, the method, system, storage medium and device for automatically marking fill-in-the-blank questions shown in the present invention have the following beneficial effects: The present invention obtains the original input information of the interface where the test questions to be marked are located, classifies the test questions to be marked according to the original input information, at least determines the grade and subject to which the test questions to be marked belong, identifies the answer area corresponding to the test questions to be marked, and the answer area corresponding to the test questions to be marked, extracts the answer content in the answer area according to the answer area, and extracts the answer content in the answer area according to the answer area, and classifies the answer content, compares the answer content with the answer content according to the marking rules corresponding to the category to which the answer content belongs, outputs the initial marking result of the test questions to be marked, uses a preset answer marking model to evaluate the initial marking result, and outputs the target marking result. The present invention solves the problem that the prior art lacks the ability to carry out overall planning and marking across disciplines and grades, and the marking ideas are relatively mechanized, which makes the marking efficiency and marking accuracy low. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which: Figure 1 Schematic diagram of the flow of the method for automatically marking fill-in-the-blank questions in one embodiment of the present invention; Figure 2 It is a structural block diagram of the automatic marking system for fill-in-the-blank questions shown in one embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings. Several embodiments of the present invention are shown in the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0020] First embodiment See also Figure 1 The first embodiment of the present invention provides a method for automatically marking fill-in-the-blank questions, the method comprising steps S10 to S50: Step S10, obtaining original input information of the interface where the test questions to be marked are located, classifying the test questions to be marked according to the original input information, and at least determining the grade and subject to which the test questions to be marked belong.

[0021] In this embodiment, the steps of obtaining original input information of the interface where the test questions to be marked are located, classifying the test questions to be marked according to the original input information, and at least determining the grade and subject to which the test questions to be marked belong include: Identify at least one interface where a test question to be reviewed is located, and obtain original input information corresponding to the test question to be reviewed in the target interface; According to the original input information, the test questions to be marked are classified by question scenarios to at least determine the grade and subject to which the test questions to be marked belong.

[0022] It should be noted here that the test questions to be marked refer to fill-in-the-blank questions that are to be automatically marked, such as fill-in-the-blank questions in primary school, junior high school and high school education, which may include subjects such as Chinese, mathematics, physics, chemistry, etc., and the interface where the test questions to be marked are located refers to the interface that displays the test questions to be marked, which is an electronic interface. Because the test paper needs to be scanned to obtain the test paper image when automatic marking is performed, the interface where the test questions to be marked are located specifically refers to the test paper image where the test questions to be marked are located.

[0023] Specifically, in this embodiment, by obtaining the original input information of the interface where the test questions to be marked are located, the original input information such as the student's name, grade, class, subject, etc., or the original information in the test paper image, the test questions to be marked can be classified into question scenarios according to the original input information to at least determine the grade and subject to which the test questions to be marked belong, which is equivalent to determining the academic ability examination scope corresponding to the test questions to be marked.

[0024] Step S20, identifying the answer area corresponding to the test question to be marked, and the answer area corresponding to the test question to be marked.

[0025] In this embodiment, OCR recognition and OCR segmentation are used to identify and extract the answer area corresponding to the test questions to be marked in the test paper image, that is, the area where students answer fill-in-the-blank questions, which is usually located in a certain area of ​​the test questions to be marked. For example, at least one answer area is interspersed in the stem of the test questions to be marked. Regardless of whether the answer area is empty, the answer area is extracted; then based on the same method, the answer area corresponding to the test questions to be marked is identified and extracted in the answer image corresponding to the standard answer.

[0026] Among them, since the answer area and the answer area are not on the same carrier, the present embodiment also establishes a correspondence between the test questions to be marked and their corresponding answer contents, including comparing the question numbers and answer numbers to output their correspondence, thereby facilitating the identification of the answer area corresponding to the test questions to be marked, and avoiding the situation where the test questions to be marked and the answer contents do not match during the identification and extraction process.

[0027] Step S30, extracting the answer content in the answer area according to the answer area, and extracting the answer content in the answer area according to the answer area, and classifying the answer content.

[0028] In this embodiment, the steps of extracting the answer content in the answer area according to the answer area, extracting the answer content in the answer area according to the answer area, and classifying the answer content include: According to the answer area, determining whether the answer area is empty, and if not, extracting the answer content in the answer area; Extracting answer content in the answer area according to the answer area; Classifying the answer content into categories according to a preset regular expression; The categories of the answer content include pure numerical value, character symbol, expression, professional terminology and text description.

[0029] In this embodiment, after the answer area and answer area of ​​the test questions to be marked are identified and extracted through OCR recognition and segmentation, the answer content in the answer area is extracted based on semantic recognition, that is, the answer content filled in by the students is identified, and the answer content in the answer area is extracted based on semantic recognition, that is, the standard answer content corresponding to the test questions to be marked is identified. Finally, the answer content is classified to clarify the category to which the answer content belongs.

[0030] In a more specific example, taking mathematics as an example, the answers to fill-in-the-blank questions include pure numerical values ​​(such as 12, 2.5, 1 / 2, etc.), character symbols (such as AB, b, a^2, f(x), etc.), expressions (such as 2a+b, a^2-b^2, y=3x+2, etc.), professional terms (such as projection theorem, Pythagorean theorem, etc.) and text descriptions (such as straight line l is perpendicular to plane a, draw a straight line through point A parallel to straight line CD, etc.).

[0031] It should also be noted that the acceptable description methods corresponding to the answers may differ in different grades or different learning stages of the same grade. For example, in elementary school or the initial stage of elementary school, half of 1 is only allowed to be written as 0.5 in numerical form, and cannot be written as 1 / 2 in fraction form. If the answer is 1 / 2 in fraction form, it is judged to be wrong. In junior high school and above, if the answer is 0.5, it is judged to be correct regardless of whether the answer is written as 0.5 in numerical form or 1 / 2 in fraction form. Therefore, in this embodiment, it is necessary to classify the answer content accordingly in combination with the original input information, so as to clarify the corresponding marking rules.

[0032] Therefore, when the category of the answer content is a professional terminology category or a text description category, the step of classifying the answer content further includes: A pre-trained deep neural network is used to classify the answer content to determine the chapter features to which the answer content belongs.

[0033] Step S40, comparing the answer content with the answer content according to the marking rule corresponding to the category to which the answer content belongs, and outputting the initial marking result of the test question to be marked.

[0034] In this embodiment, the step of comparing the answer content with the answer content according to the review rules corresponding to the category to which the answer content belongs, and outputting the initial review result of the test question to be reviewed includes: Comparing the answer content with the answer content according to the evaluation criteria corresponding to the category to which the answer content belongs; If the answer content is exactly the same as the answer content, then the answer content is judged to be correct; If the answer content is not completely the same as the answer content, the answer content is converted to obtain equivalent content, and the equivalent content is compared with the answer content to determine whether they are the same; According to the judgment result, the initial marking result corresponding to the test question to be marked is output.

[0035] In this embodiment, after classifying the answer content and determining the category to which it belongs, such as pure numerical category or expression category, the marking rule corresponding to the category to which the answer content belongs will be called first, and the response content will be compared with the answer content based on the marking rule, and then the initial marking result of the test question to be marked will be output.

[0036] It is easy to understand that if the content of the answer is exactly the same as the answer content, the answer is judged to be correct, otherwise it is judged to be wrong; if the content of the answer is not exactly the same as the answer content, a further equivalence judgment is performed.

[0037] Among them, the equivalence judgment is specifically achieved by using symbolic operation tools (such as sympy, which has a method for symbolic equivalence operation), such as 1 / 2 is equivalent to 0.5, and radians and angles are also equivalent in conversion; if it is an expression type, then it is also equivalent if it conforms to the commutative law, associative law and other laws of expression operation (such as 2a+b is equivalent to b+2a); and if it involves whether the semantic expression is equivalent, it is necessary to use the language model of deep learning to judge, and at the same time, it is necessary to rely on the professional terminology knowledge base of the subject to distinguish which ones require professional terms for precise expression and which ones can accept diversified semantic expressions to give corresponding judgment results, and the method shown in this embodiment enters step S50.

[0038] Step S50: Use a preset answer review model to evaluate the initial review result and output a target review result.

[0039] In this embodiment, the step of using a preset answer review model to evaluate the initial review result and outputting a target review result includes: Importing the original input information into a preset answer review model, so that the answer review model performs semantic understanding based on the original input information, and screens out review criteria corresponding to the original input information in a database; According to the review standard, the initial review result and / or the answer content corresponding to the professional terminology category and the text description category are reviewed, and the target review result is output.

[0040] Specifically, after obtaining the initial marking results, the planning capability of the answer marking model is utilized to make a final evaluation and judgment on the initial marking results outputted in step S40 according to the grade, subject category, and question category of the test questions to be marked.

[0041] Among them, for the unique answer category, the initial review result output by step S40 can be directly selected as the final result, while for the answer categories of expression equivalence and semantic equivalence, the answer review model will give different judgment results according to the characteristics and requirements of different grades and subjects.

[0042] In addition, before the final review, the basic information of the test needs to be provided to the answer review model. The LLM module in the answer review model will use its semantic understanding ability to select a set of appropriate review standards based on the subject, grade, learning progress and even question information of the test. In the equivalent judgment of step S40, there is often the influence of conditions. For example, in different grades or different stages of learning, the result requirements for the same question are different; for example, corresponding to the learning stage of radians, the answer must be a radian value answer. If an angle value answer is given, it is judged as wrong, but in the comprehensive examination stage, it is judged as correct.

[0043] In addition, the answer review model can also judge the effectiveness of answers in semantically equivalent categories based on the learning stage they belong to, thereby giving more accurate results.

[0044] It should also be noted that the answer review model used for the final answer review in this embodiment is the LLM model (Large Language Model).

[0045] Compared with the prior art, the automatic marking method for fill-in-the-blank questions of this embodiment has the following beneficial effects: The method shown in this embodiment obtains the original input information of the interface where the test questions to be marked are located, classifies the test questions to be marked according to the original input information, at least determines the grade and subject to which the test questions to be marked belong, identifies the answer area corresponding to the test questions to be marked, and the answer area corresponding to the test questions to be marked, extracts the answer content in the answer area according to the answer area, and extracts the answer content in the answer area according to the answer area, and classifies the answer content, compares the answer content with the answer content according to the marking rules corresponding to the category to which the answer content belongs, outputs the initial marking results of the test questions to be marked, uses a preset answer marking model to evaluate the initial marking results, and outputs the target marking results. The present invention solves the problem that the prior art lacks the ability to carry out overall planning and marking across disciplines and grades, and the marking ideas are relatively mechanized, which makes the marking efficiency and marking accuracy low.

[0046] Second embodiment See also Figure 2 The second embodiment of the present invention provides a fill-in-the-blank question automatic marking system, which is applied to the method described in the first embodiment above. The system includes: a first classification module 10, a region identification module 20, a second classification module 30, a result output module 40 and a result evaluation module 50.

[0047] The first classification module 10 is used to obtain the original input information of the interface where the test questions to be marked are located, and classify the test questions to be marked according to the original input information, and at least determine the grade and subject to which the test questions to be marked belong; An area identification module 20, used to identify the answer area corresponding to the test question to be marked, and the answer area corresponding to the test question to be marked; A second classification module 30 is used to extract the answer content in the answer area according to the answer area, and to extract the answer content in the answer area according to the answer area, and to classify the answer content; A result output module 40 is used to compare the answer content with the answer content according to the review rules corresponding to the category to which the answer content belongs, and output the initial review result of the test question to be reviewed; The result evaluation module 50 is used to evaluate the initial review result using a preset answer review model and output a target review result.

[0048] Compared with the prior art, the automatic marking system for fill-in-the-blank questions of this embodiment has the following beneficial effects: The system shown in this embodiment obtains the original input information of the interface where the test questions to be marked are located, classifies the test questions to be marked according to the original input information, at least determines the grade and subject to which the test questions to be marked belong, identifies the answer area corresponding to the test questions to be marked, and the answer area corresponding to the test questions to be marked, extracts the answer content in the answer area according to the answer area, and extracts the answer content in the answer area according to the answer area, and classifies the answer content, compares the answer content with the answer content according to the marking rules corresponding to the category to which the answer content belongs, outputs the initial marking results of the test questions to be marked, uses a preset answer marking model to evaluate the initial marking results, and outputs the target marking results. The present invention solves the problem that the prior art lacks the ability to carry out overall planning and marking across disciplines and grades, and the marking ideas are relatively mechanized, which makes the marking efficiency and marking accuracy low.

[0049] Third embodiment A third embodiment of the present invention provides a readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method described in the above embodiment is implemented.

[0050] Fourth embodiment A fourth embodiment of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor implements the method described in the above embodiments when executing the computer program.

[0051] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples" or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0052] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for automatically marking fill-in-the-blank questions, characterized in that: The method comprises: Acquire original input information of the interface where the test questions to be reviewed are located, classify the test questions to be reviewed according to the original input information, and at least determine the grade and subject to which the test questions to be reviewed belong; Identify the answer area corresponding to the test question to be marked, and the answer area corresponding to the test question to be marked; Extracting answer content in the answer area according to the answer area, and extracting answer content in the answer area according to the answer area, and classifying the answer content; According to the review rules corresponding to the category to which the answer content belongs, the answer content is compared with the answer content, and an initial review result of the test question to be reviewed is output; Using a preset answer review model to evaluate the initial review result and output a target review result; The categories of the answer content include pure numerical value, character symbol, expression, professional terminology and text description; Among them, the step of comparing the answer content with the answer content according to the review rules corresponding to the category to which the answer content belongs, and outputting the initial review result of the test question to be reviewed includes: Comparing the answer content with the answer content according to the evaluation criteria corresponding to the category to which the answer content belongs; If the answer content is exactly the same as the answer content, then the answer content is judged to be correct; If the answer content is not completely the same as the answer content, the answer content is converted to obtain equivalent content, and the equivalent content is compared with the answer content to determine whether they are the same; Outputting the initial marking result corresponding to the test question to be marked according to the judgment result; The step of using a preset answer review model to evaluate the initial review result and outputting a target review result includes: Importing the original input information into a preset answer review model, so that the answer review model performs semantic understanding based on the original input information, and screens out review criteria corresponding to the original input information in a database; According to the review standard, the initial review result and / or the answer content corresponding to the professional terminology category and the text description category are reviewed, and the target review result is output.

2. The method for automatically marking fill-in-the-blank questions according to claim 1, characterized in that: The steps of obtaining original input information of the interface where the test questions to be reviewed are located, classifying the test questions to be reviewed according to the original input information, and at least determining the grade and subject to which the test questions to be reviewed belong, include: Identify at least one interface where a test question to be reviewed is located, and obtain original input information corresponding to the test question to be reviewed in the target interface; According to the original input information, the test questions to be marked are classified by question scenarios to at least determine the grade and subject to which the test questions to be marked belong.

3. The method for automatically marking fill-in-the-blank questions according to claim 1, characterized in that: The steps of extracting answer content in the answer area according to the answer area, extracting answer content in the answer area according to the answer area, and classifying the answer content include: According to the answer area, determining whether the answer area is empty, and if not, extracting the answer content in the answer area; Extracting answer content in the answer area according to the answer area; The answer content is classified into forms according to a preset regular expression.

4. The method for automatically marking fill-in-the-blank questions according to claim 3 is characterized in that: When the category of the answer content is a professional terminology category or a text description category, the step of classifying the answer content further includes: A pre-trained deep neural network is used to classify the answer content to determine the chapter features to which the answer content belongs.

5. The method for automatically marking fill-in-the-blank questions according to any one of claims 1 to 4, characterized in that: The answer review model is the LLM model.

6. A fill-in-the-blank question automatic marking system, characterized in that: The method applied to any one of claims 1 to 5, wherein the system comprises: A first classification module is used to obtain original input information of the interface where the test questions to be reviewed are located, and classify the test questions to be reviewed according to the original input information, and at least determine the grade and subject to which the test questions to be reviewed belong; An area identification module, used to identify the answer area corresponding to the test question to be marked, and the answer area corresponding to the test question to be marked; A second classification module is used to extract the answer content in the answer area according to the answer area, and to extract the answer content in the answer area according to the answer area, and to classify the answer content; A result output module, used for comparing the answer content with the answer content according to the review rules corresponding to the category to which the answer content belongs, and outputting the initial review result of the test question to be reviewed; The result evaluation module is used to evaluate the initial review result by using a preset answer review model and output a target review result.

7. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of being executed on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

Images