Intelligent test paper composition method and device

Abstract

The invention discloses an intelligent test paper composition method and device, and belongs to the technical field of image or video recognition or understanding. The intelligent test paper composition method comprises the following steps: collecting original data of various questions, and carrying out standardization processing on the collected data to form a standard question bank; wherein the standardization processing comprises adding applicable teaching materials, applicable grades, question types, first-level knowledge point labels, second-level knowledge point labels, difficulty coefficients and score information to the original data; pre-de-duplication processing is carried out on questions in a standard question bank, then the standard question bank is divided into a plurality of first-level question banks according to first-level knowledge point labels, and the first-level question banks are divided into a plurality of second-level question banks according to second-level knowledge point labels; and screening questions from the standard question bank according to set test paper composition information. And when the repeated questions belong to the second-level question bank related to the core knowledge point, replacing the repeated questions with another question in the same second-level question bank, thereby ensuring the accuracy of the test paper composition again.

Description

Technical Field

[0001] The present invention belongs to the technical field of image or video recognition or understanding; in particular, it relates to an intelligent test paper composition method and device. Background Art

[0002] In the field of education, test preparation is a critical and complex task. Traditional manual test preparation has significant limitations: teachers spend a significant amount of time and effort selecting test questions, assigning scores, and balancing difficulty. This is followed by grading, data analysis, and archiving. This overall process is inefficient and susceptible to subjective factors, potentially leading to illogical test structure and biased assessment results.

[0003] Chinese patent CN114005116A discloses a test paper assembly method, device, and storage medium based on test paper information. The test paper assembly method based on test paper information includes: obtaining test paper attribute information of the test paper to be assembled; retrieving a similar test paper set in a test paper library based on the test paper attribute information; analyzing the similar test paper set to obtain test paper structure information, test question knowledge point information, and test question difficulty information of the test paper to be assembled; retrieving similar test questions in the test question library using the test paper structure information, test question knowledge point information, and test question difficulty information as screening conditions; and assembling the retrieved similar test questions into a test paper. The test paper assembly method of the present invention realizes the function of automatically assembling test papers according to test paper assembly requirements, meets the test paper assembly requirements, and greatly improves the efficiency of test paper assembly.

[0004] However, the above patent document cannot achieve accurate deduplication during the paper compilation process with less computing power. Summary of the Invention

[0005] The purpose of the present disclosure is to provide an intelligent test paper composition method and device, which can solve at least some of the above technical problems.

[0006] A first aspect of an embodiment of the present disclosure provides an intelligent test paper composition method, the intelligent test paper composition method comprising:

[0007] Collecting raw data of various questions and standardizing the collected data to form a standard question bank; wherein the standardization includes adding applicable textbooks, applicable grades, question types, first-level knowledge point labels, second-level knowledge point labels, difficulty coefficients, and score information to the raw data;

[0008] Performing pre-duplicate removal on the questions in the standard question bank, and then dividing the standard question bank into multiple first-level question banks according to the first-level knowledge point tags, and dividing the first-level question bank into multiple second-level question banks according to the second-level knowledge point tags;

[0009] According to the set test information, questions are selected from the standard question bank. The specific process is as follows:

[0010] According to the total score of the test paper and the requirements for question type configuration, reasonably allocate the number of each question type and the score range of each question;

[0011] According to the preset difficulty distribution ratio, select questions with corresponding difficulty coefficients;

[0012] Selecting at least one question from each of the first-level question banks associated with the test paper information, and giving priority to selecting from the second-level question banks related to core knowledge points;

[0013] After the preliminary test paper compilation is completed, the screening questions will be subjected to final deduplication processing. During the final deduplication process, if it is determined that the screening questions are repeated, priority will be given to determining whether the repeated questions belong to the secondary question bank related to the core knowledge points. If it is determined that the repeated questions belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same secondary question bank; if it is determined that the repeated questions do not belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same primary question bank.

[0014] Optionally, the first-level knowledge point tag includes a plurality of associated second-level knowledge point tags.

[0015] Optionally, the pre-deduplicate processing includes a first-level detection means, which detects whether the question type and the second-level knowledge point label of the screening question are repeated. When the question type and the second-level knowledge point label of two questions in the screening question are exactly the same, the first-level detection means determines that the screening question has repeated knowledge points.

[0016] Optionally, the pre-deduplicate processing also includes a secondary detection means, which determines whether the similarity between the answers and the problem-solving steps of the screening questions exceeds a first set threshold; when the similarity between the answers and the problem-solving steps of two questions in the screening questions exceeds the first set threshold, the secondary detection means determines that the screening questions contain repeated knowledge points; wherein, when the first-level detection means determines that the screening questions contain repeated knowledge points, the second-level detection means is entered; when the first-level detection means determines that the screening questions do not contain repeated knowledge points, the second-level detection means is not entered.

[0017] Optionally, the pre-deduplicate processing also includes a three-level detection means, which determines whether the semantic similarity of the screening questions exceeds a second set threshold; when the semantic similarity between two questions in the screening questions exceeds the second set threshold, the three-level detection means determines that repeated knowledge points appear in the screening questions; wherein, when the second-level detection means determines that repeated knowledge points appear in the screening questions, the third-level detection means is entered; when the second-level detection means determines that no repeated knowledge points appear in the screening questions, the third-level detection means is not entered.

[0018] Optionally, the third-level detection means determines the semantic similarity of the screening questions by using BERT. The specific steps of determining the semantic similarity of the screening questions by using BERT include:

[0019] Clean up the screening questions: remove irrelevant content and unify the format;

[0020] Word segmentation coding: Split the sentences in the screening questions into words and encode each word;

[0021] Generate sentence fingerprints: Let BERT process the encoded text to obtain feature vectors for all words, and then average all word vectors to obtain the sentence fingerprint of the screening question;

[0022] Comparison similarity: Calculate the cosine similarity between the sentence fingerprints of two screening questions.

[0023] Optionally, the final deduplication process is performed by judging whether the semantic similarity of the screening questions exceeds a third set threshold.

[0024] A second aspect of the embodiments of the present disclosure provides an electronic device, characterized in that:

[0025] including a processor and a memory;

[0026] The memory is used to store programs;

[0027] The processor executes the program to implement the method described above.

[0028] A third aspect of the embodiments of the present disclosure provides a computer-readable storage medium, characterized in that:

[0029] The storage medium stores a program, and the program is executed by the processor to implement the method described above.

[0030] A fourth aspect of the embodiments of the present disclosure provides an intelligent test paper assembly device, the intelligent test paper assembly device comprising:

[0031] The data collection module is used to collect raw data of various questions and standardize the collected data to form a standard question bank; wherein the standardization processing includes adding applicable textbooks, applicable grades, question types, first-level knowledge point labels, second-level knowledge point labels, difficulty coefficients, and score information to the raw data;

[0032] A pre-processing module is used to pre-de-duplicate questions in the standard question bank, and then divide the standard question bank into multiple first-level question banks according to the first-level knowledge point tags, and divide the first-level question bank into multiple second-level question banks according to the second-level knowledge point tags;

[0033] The test paper composition rule setting module provides a human-computer interaction interface for users to input test paper composition information, and parses and stores the input information;

[0034] The question screening and matching module selects the qualified questions from the question bank according to the set test information;

[0035] The test paper generation and adjustment module is used to reasonably allocate the number of each question type and the score range of each question according to the total score of the test paper and the question type configuration requirements;

[0036] Used to filter questions with corresponding difficulty coefficients according to the preset difficulty distribution ratio;

[0037] Used to select at least one question from each of the first-level question banks associated with the test paper information, and give priority to selecting from the second-level question banks related to core knowledge points;

[0038] It is used to perform final deduplication processing on the screening questions after the preliminary completion of the test paper. During the final deduplication process, if it is determined that the screening questions are repeated, priority will be given to determining whether the repeated questions belong to the secondary question bank related to the core knowledge points. If it is determined that the repeated questions belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same secondary question bank; if it is determined that the repeated questions do not belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same primary question bank.

[0039] The technical effect of the disclosed embodiment is as follows: by collecting and standardizing question data, the standard question bank is divided into a first-level question bank and a second-level question bank, wherein the first-level question bank has a broad knowledge scope, and the second-level question bank has precise knowledge points. In the process of test paper compilation, questions are first screened from the first-level question bank associated with the test paper information to ensure the comprehensiveness of the test points of the test paper, and then the questions in the test paper are finally deduplicated. In the process of final deduplication, it is first determined whether the repeated questions belong to the second-level question bank related to the core knowledge points, and when the repeated questions belong to the second-level question bank related to the core knowledge points, the repeated questions are replaced with another question in the same second-level question bank, thereby ensuring the accuracy of the test paper compilation again. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A flowchart of an intelligent test paper composition method according to an embodiment of the present disclosure is shown;

[0041] Figure 2 The figure is a flowchart of selecting questions from a standard question bank in one embodiment of the present disclosure. DETAILED DESCRIPTION

[0042] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of devices consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0043] The terms used in this disclosure are for the purpose of describing specific embodiments only and are not intended to limit the disclosure. Unless otherwise defined, the technical or scientific terms used in this disclosure should have the ordinary meaning understood by a person of ordinary skill in the art to which this disclosure belongs. The words "first," "second," and similar terms used in this disclosure and the claims do not indicate any order, quantity, or importance, but are simply used to distinguish different components. Similarly, words such as "a" or "an" do not indicate a quantitative limitation, but rather indicate the presence of at least one. If only "one" is referred to, it will be separately stated. "Multiple" or "several" means two or more. Unless otherwise specified, words such as "front," "rear," "lower," and / or "upper" are for ease of description only and are not limited to a single position or spatial orientation. Words such as "include" or "comprises" mean that the elements or objects listed before "include" or "comprises" include the elements or objects listed after "include" or "comprises" and their equivalents, and do not exclude other elements or objects. "Connected" or "connected" and similar words are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. As used in this disclosure and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0044] like Figure 1 and Figure 2 As shown, the first aspect of the embodiment of the present disclosure provides an intelligent test paper composition method, the intelligent test paper composition method comprising:

[0045] Raw data from various types of questions is collected and standardized to form a standard question bank. This standardization process includes appending information such as applicable textbooks, applicable grade levels, question types, primary and secondary knowledge point labels, difficulty coefficients, and score values ​​to the raw data. Standardization eliminates discrepancies between data from different sources, ensuring data consistency and comparability. For example, questions from different textbooks may use different wording or knowledge point labels. Standardization can unify these questions into a common framework, facilitating subsequent screening and management.

[0046] The questions in the standard question bank are pre-deduplicated, and then the standard question bank is divided into multiple first-level question banks according to the first-level knowledge point labels, and the first-level question bank is divided into multiple second-level question banks according to the second-level knowledge point labels; the pre-deduplication process quickly identifies redundant questions by detecting the duplication of question types and knowledge point labels. For example, if two questions are exactly the same in type and knowledge point label, they can be judged as duplicate questions. This preliminary deduplication can significantly reduce redundant data in the question bank and improve the efficiency of subsequent screening, thereby improving the purity of the question bank, reducing the waste of computing resources, and ensuring the efficiency of the subsequent test preparation process. The sub-library management can classify questions according to the hierarchical structure of knowledge points, making question screening more efficient and accurate. For example, the first-level knowledge point labels can be "solid geometry", "plane geometry", etc., and the second-level knowledge point labels can be "triangle", "parallelogram", etc. Through this hierarchical management, the test preparer can quickly locate the questions of specific knowledge points.

[0047] According to the set test information, questions are selected from the standard question bank. The specific process is as follows:

[0048] Based on the total score and question type requirements, the number of each question type and the score range for each question are reasonably allocated. The screening process calculates the number and score range for each question type by analyzing the total score and question type requirements. For example, if the total score is 100 points, multiple-choice questions account for 50 points, fill-in-the-blank questions account for 30 points, and essay questions account for 20 points, then the question bank will be screened for eligible questions based on these requirements. This reasonable allocation ensures a well-structured and comprehensive exam.

[0049] According to the preset difficulty distribution ratio, select questions with corresponding difficulty coefficients;

[0050] At least one question is selected from each of the primary question banks associated with the test-taking information, with priority given to selecting from the secondary question banks related to core knowledge points. Core knowledge points are typically key areas of instruction and examinations, and prioritizing selection of questions related to these points ensures the relevance and effectiveness of the test paper. For example, if the core knowledge point is "parallelogram," questions are prioritized from the secondary question bank related to "parallelogram." This ensures that the test paper highlights key points, covers core knowledge points, and improves the quality of the test paper and teaching effectiveness.

[0051] After the initial test paper is assembled, the screening questions will be subjected to a final deduplication process. During this final deduplication process, if duplicates are detected in the screening questions, priority will be given to determining whether the duplicate questions belong to the secondary question bank related to core knowledge points. If the duplicate questions are determined to belong to the secondary question bank related to core knowledge points, the duplicate questions will be replaced with another question from the same secondary question bank; if the duplicate questions are determined not to belong to the secondary question bank related to core knowledge points, the duplicate questions will be replaced with another question from the same primary question bank. After the initial test paper is assembled, deduplication will be performed again to ensure that there are no duplicate questions in the test paper. At the same time, questions related to core knowledge points will be retained first to avoid affecting the quality of the test paper.

[0052] To sum up, by collecting and standardizing question data, the standard question bank is divided into a first-level question bank and a second-level question bank. Among them, the first-level question bank has a broad knowledge scope, and the second-level question bank has precise knowledge points. In the process of test paper compilation, questions are first screened from the first-level question bank associated with the test paper information to ensure the comprehensiveness of the test points, and then the questions in the test paper are finally deduplicated. In the process of final deduplication, it is prioritized to determine whether the repeated questions belong to the second-level question bank related to the core knowledge points, and when the repeated questions belong to the second-level question bank related to the core knowledge points, the repeated questions are replaced with another question in the same second-level question bank, thereby ensuring the accuracy of the test paper compilation again.

[0053] As an optional embodiment, the first-level knowledge point tag includes multiple associated second-level knowledge point tags. The first-level knowledge point tag includes multiple second-level knowledge point tags, making the hierarchical relationship between knowledge points clearer and facilitating accurate matching of user needs when screening questions.

[0054] As an optional implementation method, the pre-deduplication processing includes a first-level detection means, which detects whether the question type and the second-level knowledge point label of the screening question are repeated. When the question type and the second-level knowledge point label of two questions in the screening question are exactly the same, the first-level detection means determines that the screening question has repeated knowledge points.

[0055] The pre-deduplication processing also includes a secondary detection means, which determines whether the similarity between the answers and the solution steps of the screening questions exceeds a first set threshold; when the similarity between the answers and the solution steps of two questions in the screening questions exceeds the first set threshold, the secondary detection means determines that the screening questions contain repeated knowledge points; wherein, when the first-level detection means determines that the screening questions contain repeated knowledge points, the second-level detection means is entered; when the first-level detection means determines that the screening questions do not contain repeated knowledge points, the second-level detection means is not entered.

[0056] The pre-deduplication processing also includes a three-level detection means, which determines whether the semantic similarity of the screening questions exceeds a second set threshold; when the semantic similarity between two questions in the screening questions exceeds the second set threshold, the three-level detection means determines that repeated knowledge points appear in the screening questions; wherein, when the second-level detection means determines that repeated knowledge points appear in the screening questions, the third-level detection means is entered; when the second-level detection means determines that no repeated knowledge points appear in the screening questions, the third-level detection means is not entered.

[0057] The pre-deduplicate processing first determines whether the question type and the secondary knowledge point label are repeated, and then determines whether the similarity between the answer to the screening question and the problem-solving steps exceeds the first set threshold, and finally determines whether the semantic similarity of the screening question exceeds the second set threshold. The above judgment order allows the processor to use only a small amount of computing power to filter out most of the data in the original data through the first-level detection method, and then use the second-level and third-level detection methods that occupy more computing power to process the small part of the data filtered out by the first-level detection method, so that the general processor can deduplicate the entire original data, thereby reducing costs.

[0058] As an optional implementation, the three-level detection means determines the semantic similarity of the screening questions by using BERT. The specific steps of determining the semantic similarity of the screening questions by using BERT include:

[0059] Clean the screening questions: Remove irrelevant content and unify the format. Questions may contain irrelevant content (such as redundant symbols, inconsistent formatting, etc.), which can affect subsequent semantic analysis. Data cleaning can eliminate these interference factors and improve the accuracy of semantic similarity judgment.

[0060] Word segmentation and encoding: Sentences in the screening questions are broken down into words and each word is encoded. Word segmentation is a fundamental step in natural language processing. By breaking sentences down into words, we can better understand their structure and semantics. Encoding converts the words into a form that can be processed by computers, facilitating subsequent feature extraction and analysis.

[0061] Generating sentence fingerprints: BERT processes the encoded text to obtain feature vectors for all words. These vectors are then averaged to form the sentence fingerprint for the screening question. BERT is a pretrained language model that understands the semantics of natural language. Processing the encoded text with BERT extracts the semantic feature vector for each word. Averaging these vectors yields a semantic feature representation of the entire sentence (the sentence fingerprint), which provides the basis for subsequent semantic similarity calculations.

[0062] Comparison Similarity: Calculate the cosine similarity between the sentence fingerprints of two screening questions. Cosine similarity is a commonly used similarity metric that calculates the cosine of the angle between two vectors. By calculating the cosine similarity between two sentence fingerprints, we can quantify their semantic similarity. If the similarity exceeds a set threshold, the two questions can be considered semantically duplicated.

[0063] As an optional implementation, the final deduplication process is performed by determining whether the semantic similarity of the screened questions exceeds a third set threshold. This is because in the pre-processing deduplication process, in order to save computing power, the pre-processing deduplication process actually only removes duplicate questions of the same type. However, there are also different types of duplicate questions in the standard question bank. The final deduplication process and the pre-processing deduplication process cooperate with each other to achieve the effect of removing duplicate questions of different types with less computing power.

[0064] A second aspect of the embodiments of the present disclosure provides an electronic device, characterized in that:

[0065] including a processor and a memory;

[0066] The memory is used to store programs;

[0067] The processor executes the program to implement the method described above.

[0068] A third aspect of the embodiments of the present disclosure provides a computer-readable storage medium, characterized in that:

[0069] The storage medium stores a program, and the program is executed by the processor to implement the method described above.

[0070] A fourth aspect of the embodiments of the present disclosure provides an intelligent test paper assembly device, the intelligent test paper assembly device comprising:

[0071] The data collection module is used to collect raw data of various questions and standardize the collected data to form a standard question bank; wherein the standardization processing includes adding applicable textbooks, applicable grades, question types, first-level knowledge point labels, second-level knowledge point labels, difficulty coefficients, and score information to the raw data;

[0072] A pre-processing module is used to pre-de-duplicate questions in the standard question bank, and then divide the standard question bank into multiple first-level question banks according to the first-level knowledge point tags, and divide the first-level question bank into multiple second-level question banks according to the second-level knowledge point tags;

[0073] The test paper composition rule setting module provides a human-computer interaction interface for users to input test paper composition information, and parses and stores the input information;

[0074] The question screening and matching module selects the qualified questions from the question bank according to the set test information;

[0075] The test paper generation and adjustment module is used to reasonably allocate the number of each question type and the score range of each question according to the total score of the test paper and the question type configuration requirements;

[0076] Used to filter questions with corresponding difficulty coefficients according to the preset difficulty distribution ratio;

[0077] Used to select at least one question from each of the first-level question banks associated with the test paper information, and give priority to selecting from the second-level question banks related to core knowledge points;

[0078] It is used to perform final deduplication processing on the screening questions after the preliminary completion of the test paper. During the final deduplication process, if it is determined that the screening questions are repeated, priority will be given to determining whether the repeated questions belong to the secondary question bank related to the core knowledge points. If it is determined that the repeated questions belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same secondary question bank; if it is determined that the repeated questions do not belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same primary question bank.

[0079] The technical effect of the disclosed embodiment is as follows: by collecting and standardizing question data, the standard question bank is divided into a first-level question bank and a second-level question bank, wherein the first-level question bank has a broad knowledge scope, and the second-level question bank has precise knowledge points. In the process of test paper compilation, questions are first screened from the first-level question bank associated with the test paper information to ensure the comprehensiveness of the test points of the test paper, and then the questions in the test paper are finally deduplicated. In the process of final deduplication, it is first determined whether the repeated questions belong to the second-level question bank related to the core knowledge points, and when the repeated questions belong to the second-level question bank related to the core knowledge points, the repeated questions are replaced with another question in the same second-level question bank, thereby ensuring the accuracy of the test paper compilation again.

[0080] The above are only preferred embodiments of the present disclosure and are not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. An intelligent test paper composition method, characterized in that: The intelligent test paper composition method includes: Collecting raw data of various questions and standardizing the collected data to form a standard question bank; wherein the standardization includes adding applicable textbooks, applicable grades, question types, first-level knowledge point labels, second-level knowledge point labels, difficulty coefficients, and score information to the raw data; Performing pre-duplicate removal on the questions in the standard question bank, and then dividing the standard question bank into multiple first-level question banks according to the first-level knowledge point tags, and dividing the first-level question bank into multiple second-level question banks according to the second-level knowledge point tags; According to the set test information, questions are selected from the standard question bank. The specific process is as follows: According to the total score of the test paper and the requirements for question type configuration, reasonably allocate the number of each question type and the score range of each question; According to the preset difficulty distribution ratio, select questions with corresponding difficulty coefficients; Selecting at least one question from each of the first-level question banks associated with the test paper information, and giving priority to selecting from the second-level question banks related to core knowledge points; After the preliminary test paper compilation is completed, the screening questions will be subjected to final deduplication processing. During the final deduplication process, if it is determined that the screening questions are repeated, priority will be given to determining whether the repeated questions belong to the secondary question bank related to the core knowledge points. If it is determined that the repeated questions belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same secondary question bank; if it is determined that the repeated questions do not belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same primary question bank.

2. The intelligent test paper composition method according to claim 1, characterized in that: The first-level knowledge point tag includes a plurality of associated second-level knowledge point tags.

3. The intelligent test paper composition method according to claim 1, characterized in that: The pre-deduplication processing includes a first-level detection means, which detects whether the question type and the second-level knowledge point label of the screening question are repeated. When the question type and the second-level knowledge point label of two questions in the screening question are exactly the same, the first-level detection means determines that there are repeated knowledge points in the screening question.

4. The intelligent test paper composition method according to claim 3, characterized in that: The pre-deduplication processing also includes a secondary detection means, which determines whether the similarity between the answers and the solution steps of the screening questions exceeds a first set threshold; when the similarity between the answers and the solution steps of two questions in the screening questions exceeds the first set threshold, the secondary detection means determines that the screening questions contain repeated knowledge points; wherein, when the first-level detection means determines that the screening questions contain repeated knowledge points, the second-level detection means is entered; when the first-level detection means determines that the screening questions do not contain repeated knowledge points, the second-level detection means is not entered.

5. The intelligent test paper composition method according to claim 4, characterized in that: The pre-deduplication processing also includes a three-level detection means, which determines whether the semantic similarity of the screening questions exceeds a second set threshold; when the semantic similarity between two questions in the screening questions exceeds the second set threshold, the three-level detection means determines that repeated knowledge points appear in the screening questions; wherein, when the second-level detection means determines that repeated knowledge points appear in the screening questions, the third-level detection means is entered; when the second-level detection means determines that no repeated knowledge points appear in the screening questions, the third-level detection means is not entered.

6. The intelligent test paper composition method according to claim 5, characterized in that: The third-level detection means uses BERT to determine the semantic similarity of the screening questions. The specific steps of determining the semantic similarity of the screening questions using BERT include: Clean up the screening questions: remove irrelevant content and unify the format; Word segmentation coding: Split the sentences in the screening questions into words and encode each word; Generate sentence fingerprints: Let BERT process the encoded text to obtain feature vectors for all words, and then average all word vectors to obtain the sentence fingerprint of the screening question; Comparison similarity: Calculate the cosine similarity between the sentence fingerprints of two screening questions.

7. The intelligent test paper composition method according to claim 1, characterized in that: The final deduplication process is performed by determining whether the semantic similarity of the screening questions exceeds a third set threshold.

8. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that The storage medium stores a program, and the program is executed by a processor to implement the method according to any one of claims 1 to 7.

10. An intelligent test paper assembling device, characterized in that: The intelligent test paper assembly device includes: The data collection module is used to collect raw data of various questions and standardize the collected data to form a standard question bank; wherein the standardization processing includes adding applicable textbooks, applicable grades, question types, first-level knowledge point labels, second-level knowledge point labels, difficulty coefficients, and score information to the raw data; A pre-processing module is used to pre-de-duplicate questions in the standard question bank, and then divide the standard question bank into multiple first-level question banks according to the first-level knowledge point tags, and divide the first-level question bank into multiple second-level question banks according to the second-level knowledge point tags; The test paper composition rule setting module provides a human-computer interaction interface for users to input test paper composition information, and parses and stores the input information; The question screening and matching module selects the qualified questions from the question bank according to the set test information; The test paper generation and adjustment module is used to reasonably allocate the number of each question type and the score range of each question according to the total score of the test paper and the question type configuration requirements; Used to filter questions with corresponding difficulty coefficients according to the preset difficulty distribution ratio; Used to select at least one question from each of the first-level question banks associated with the test paper information, and give priority to selecting from the second-level question banks related to core knowledge points; It is used to perform final deduplication processing on the screening questions after the preliminary completion of the test paper. During the final deduplication process, if it is determined that the screening questions are repeated, priority will be given to determining whether the repeated questions belong to the secondary question bank related to the core knowledge points. If it is determined that the repeated questions belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same secondary question bank; if it is determined that the repeated questions do not belong to the secondary question bank related to the core knowledge points, the repeated questions will be replaced with another question in the same primary question bank.

Images