Medical theory examination intelligent test paper composition method based on artificial intelligence technology

Abstract

The invention relates to the technical field of question test paper composition, and discloses a medical theory examination intelligent test paper composition method based on an artificial intelligence technology, and the method comprises the steps: question information preprocessing, question similarity calculation, and intelligent test paper composition. In the topic processing link, the topics are subjected to disassembling and structured storage according to multiple modes of text, image and case analysis, and features are extracted in combination with technologies such as keyword extraction and local feature descriptor methods; in question similarity calculation, cosine similarity and a weighted summation mode are utilized to accurately quantify the similarity of questions of different question types, and question repetition in test paper is effectively avoided; during test paper composition, screening is performed according to parameters such as preset question type structures, knowledge point coverage and difficulty levels, reasonable difficulty distribution and comprehensive knowledge point coverage are ensured in combination with a similarity threshold judgment and replacement mechanism, a replacement object is determined according to historical use times of questions, test paper composition scientificity and question bank resource utilization rate are improved, and test paper composition efficiency is improved. And the medical theory examination paper composition flow is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of question paper compilation, and particularly relates to an intelligent question paper compilation method for medical theory examinations based on artificial intelligence technology. Background Art

[0002] In the fields of medical education and talent assessment, medical theory examinations are a key link in evaluating medical students' knowledge reserves, clinical thinking, and practical abilities. The quality of the test papers directly affects the scientificity and effectiveness of the assessment results. Traditional methods for compiling medical theory examination papers rely mostly on manual operations. Compilers need to manually select questions from a vast question bank according to the examination syllabus, which not only consumes a large amount of time and energy but also makes it difficult to accurately control the question type distribution, knowledge point coverage, and difficulty gradient of the test papers, easily leading to problems such as unbalanced test paper structures, repeated questions, or missing knowledge points.

[0003] With the booming development of artificial intelligence technology, some intelligent question paper compilation methods have emerged, but there are still significant deficiencies. Existing technologies often do not fully consider the complexity and particularity of medical examinations, lack systematic classification processing and structured storage for multi-modal questions such as text-based, image-based, and case analysis-based questions, and it is difficult to accurately extract the characteristics of various types of questions, resulting in deviations in question similarity calculations and the inability of the question paper compilation results to meet diverse examination requirements. In addition, during the question paper compilation process, high-similarity questions cannot be effectively avoided, easily leading to problems such as repeated assessment content and insufficient discrimination, and it is impossible to comprehensively and objectively evaluate the medical knowledge level and comprehensive application ability of candidates. Therefore, there is an urgent need for a scientific, efficient, and intelligent question paper compilation method suitable for medical theory examinations to improve the quality of examinations and assessment efficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent question paper compilation method for medical theory examinations based on artificial intelligence technology, which solves the technical problems raised in the background art.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] An intelligent question paper compilation method for medical theory examinations based on artificial intelligence technology includes the following steps:

[0007] First step, preprocessing of question information: Decompose questions into multi-modal questions of text-based, image-based, and case analysis-based according to question types, and extract their features. Feature extraction includes:

[0008] Text feature extraction: Segment the text parts of text-based and case analysis-based questions, calculate word frequencies and inverse document frequencies, and determine the keyword set;

[0009] Image feature extraction: Divide the image parts of image-based and case analysis-based questions into blocks, extract color and texture features, and construct an image feature vector set;

[0010] Step 2, calculation of question similarity: Based on the keyword set and the feature vector set, calculate the text similarity, image similarity, and overall multi-modal question similarity respectively;

[0011] Step 3, intelligent test paper compilation: Screen questions according to the preset question type structure, knowledge point coverage, and difficulty level requirements, and optimize the combination to obtain a set of test paper questions.

[0012] As a further solution of the present invention: For text questions: directly extract the text content and represent it with a text set W = {W(i1, j1)};

[0013] where i1 = 1, 2,..., n1, n1 represents the number of text questions, j1 = 1, 2,..., m1, m1 represents the number of words in the i1-th text question; and W(i1, j1) represents the j1-th word in the i1-th text question;

[0014] For image questions: directly extract the original image data and represent it with an image set X = {X(i2, j2)};

[0015] where i2 = 1, 2,..., n2, n2 represents the number of image questions, j2 = 1, 2,..., m2, m2 represents the number of images in the i-th image question; and X(i2, j2) represents the j2-th image in the i2-th image question;

[0016] For case analysis multi-modal questions: separate the text part and the image part, and represent the text part with WX = {WX(i3, j a ), and represent the image part with XW = {X(i3, j b )};

[0017] where i3 = 1, 2,..., n3, n3 represents the number of case analysis multi-modal questions, j a = 1, 2,..., m a , m a represents the number of words in the i3-th case analysis multi-modal question; j b = 1, 2,..., m b , m b represents the number of images in the i3-th case analysis multi-modal question; and X(i3, j a ) represents the j a -th word in the i3-th case analysis multi-modal question, and X(i3, j b ) represents the j b -th image in the i3-th case analysis multi-modal question.

[0018] As a further solution of the present invention, the text feature extraction method is as follows:

[0019] For W = {W(i1, j1)} and WX = {WX(i3, j a )}, the keyword extraction technology is used to obtain text features, and the specific steps are as follows:

[0020] Step 1.2.1, Word segmentation processing:

[0021] Using the word segmentation technology in natural language processing, the text is segmented into independent words;

[0022] Step 1.2.2, Calculate word frequency:

[0023] Count the number of times each word appears in each title containing text content C(i, j);

[0024] Among them, i ∈ {i1, i3}, j ∈ {j1, j a};

[0025] Through , calculate the frequency TF(i, j) of the j-th word in the i-th document;

[0026] Among them, m ∈ {m1, m a};

[0027] Step 1.2.3, Calculate inverse document frequency:

[0028] Then select each word one by one, and count the number of documents containing the corresponding word in the titles with text content, and record it as Dj;

[0029] Then through , calculate the inverse document frequency IDF(j) of each word;

[0030] Step 1.2.4, Determine keywords:

[0031] After that, through , calculate the term frequency-inverse document frequency TF-IDF(i, j) of each word;

[0032] Then select the words with TF-IDF(i, j) higher than the corresponding preset TF-IDF threshold to form the keyword set K.

[0033] As a further solution of the present invention, the image feature extraction method is as follows:

[0034] For X = {X(i2, j2)} and XW = {X(i3, j b )}, the local feature descriptor method is used to extract image features, and the specific steps are as follows:

[0035] Step 1.3.1, Image Blocking:

[0036] Select an image and divide the image into s local regions;

[0037] Step 1.3.2, Color Feature Extraction:

[0038] Convert the image from the RGB color space to the HSV color space, and then calculate the color mean and variance of each local region:

[0039] Specifically include:

[0040] The mean HP(k) and variance HF(k) of hue H;

[0041] The mean SP(k) and variance SF(k) of saturation S;

[0042] The mean VP(k) and variance VF(k) of brightness V;

[0043] where k = 1, 2, …… s;

[0044] Step 1.3.3, Texture Feature Extraction:

[0045] Use the gray-level co-occurrence matrix to calculate the texture parameters of each local region in the image;

[0046] Specifically include:

[0047] Energy E(k), entropy H(k), contrast C(k), correlation R(k);

[0048] Step 1.3.4, Constructing Feature Vectors:

[0049] Combine the color features and texture features extracted in the local regions to form the feature vector of the local region:

[0050] That is, F(k) = [HP(k), HF(k), SP(k), SF(k), VP(k), VF(k), E(k), H(k), C(k), R(k)];

[0051] Then combine the feature vectors corresponding to all local regions included in the image to form the image feature vector set, that is, F = [F(k) = F1, F2, …… Fs].

[0052] As a further solution of the present invention: The text similarity calculation method is as follows:

[0053] Based on the cosine similarity calculation method of the keyword set, the steps are as follows:

[0054] Select two text-based topics, extract their corresponding keyword sets Ka = {Ka(g1)} and Kb = {Ka(g2)}, and convert them into binary vectors;

[0055] where g2 = 1, 2, …… v1, v1 represents the number of all keywords in the keyword set Ka, and g2 = 1, 2, …… v2; v2 represents the number of all keywords in the keyword set Kb;

[0056] where the vector length of the binary vector is the number r of all non-repeated keywords in the two texts. If a keyword appears, it is recorded as 1, and if it does not appear, it is recorded as 0;

[0057] Obtain the binary vectors: Va = [Va(f)] and Vb = [Vb(f)];

[0058] where f = 1, 2, …… r;

[0059] Through: , calculate the text similarity SK between the two text-based topics.

[0060] As a further solution of the present invention: The image similarity calculation method is as follows:

[0061] Based on the cosine similarity calculation method of the image feature vector set, the steps are as follows:

[0062] Select two image-based topics, extract their corresponding keyword sets Fa = {Fa(k)} and Fb = {Fb(k)}, and convert them into binary vectors;

[0063] Through: , calculate the image similarity SF between the two image-based topics.

[0064] As a further solution of the present invention: The similarity calculation method for case analysis multi-modal topics is as follows:

[0065] Based on the text and image similarity calculation methods, adopt the weighted summation method, and the steps are as follows:

[0066] Select two case analysis multi-modal topics, calculate the similarity SK0 of the corresponding text parts of the two according to the text similarity calculation method, and at the same time, calculate the similarity SF0 of the corresponding image parts of the two according to the image similarity calculation method;

[0067] Through: , calculate the overall similarity SZ between the two case analysis multi-modal topics;

[0068] In the formula, α K and α F are the corresponding preset weight coefficients, and α K+α F = 1。

[0069] As a further solution of the present invention: The intelligent test paper compilation method is as follows:

[0070] Step R1, extract the test paper compilation requirements:

[0071] Extract the following parameters preset according to the test purpose:

[0072] Question type structure: The total number of questions is N, and the question type distribution is the number of text-based questions Nx, the number of image-based questions Ny, and the number of case analysis questions Nz;

[0073] And it satisfies N = Nx + Ny + Nz;

[0074] Knowledge point coverage: The knowledge point set S = {S(q)}, and each knowledge point S(q) has at least e(q) questions covering it, where e(q) is a preset value;

[0075] Difficulty level: The questions are divided into three levels: low, medium, and high, and the number of questions at each level is N1, N2, and N3 in sequence, satisfying N = N1 + N2 + N3;

[0076] Step R2, question screening:

[0077] Step R2.1, question type screening:

[0078] Select text-based, image-based, and case analysis questions that meet the quantity requirements from the pre-established question bank respectively, and form question sets Tx, Ty, and Tz, and they satisfy:

[0079] ∣Tx∣ = Nx, ∣Ty∣ = Ny, ∣Tz∣ = Nz;

[0080] Among them, ∣Tx∣, ∣Ty∣, and ∣Tz∣ refer to the quantities corresponding to text-based, image-based, and case analysis questions in the question set respectively;

[0081] Step R2.2, knowledge point screening:

[0082] For each knowledge point S(q), select the questions involving this knowledge point from Tx, Ty, and Tz, and at the same time satisfy that the knowledge point S(q) has at least e(q) questions covering it;

[0083] Step R2.3, difficulty screening:

[0084] According to the preset difficulty level of the questions, select the questions that meet the difficulty quantity requirements from the screened questions, and form question sets T1, T2, and T3, and they satisfy:

[0085] ∣T1∣ = N1, ∣T2∣ = N2, ∣T3∣ = N3;

[0086] Among them, ∣T1∣, ∣T2∣, and ∣T3∣ represent the respective quantities corresponding to the questions of each difficulty level in the question set.

[0087] Step R3, question combination optimization:

[0088] Step R3.1, similarity extraction:

[0089] For the question set of the same question type, extract the similarity S0 between each pair of questions.

[0090] Among them, S0 ∈ {SK, SF, SZ};

[0091] Step R3.2, threshold judgment:

[0092] Extract the pre-set similarity threshold S0y;

[0093] Among them, S0y ∈ {SKy, SFy, SZy}; SKy represents the text threshold, SFy represents the image threshold, and SZy represents the multi-modal threshold;

[0094] [[ID=2⑤]]Between two questions in the question set of the same question type, when their similarity S0 > similarity threshold S0y, they are determined as a highly similar question pair;

[0095] Then randomly select one question from the remaining questions of the original screening conditions, that is, a question that meets the requirements of question type structure, knowledge point coverage, and difficulty level correspondence but has not been selected, and replace one question in the highly similar question pair;

[0096] Then recalculate the similarity between the replaced question and other questions until the similarity S0 of all question pairs is less than or equal to the similarity threshold S0y, and obtain the question set for test paper compilation.

[0097] As a further solution of the present invention: Among them, the question to be replaced in the highly similar question pair is determined by its historical usage times in the question bank, specifically, the question with the largest historical usage times value is replaced.

[0098] Advantages of the present invention:

[0099] Efficient question processing and feature extraction: By carefully disassembling and structurally storing medical examination questions in multiple modalities such as text, images, and case analysis, and cooperating with advanced text feature extraction technologies (such as keyword extraction) and image feature extraction methods (such as local feature descriptor methods), the core information of the questions can be accurately obtained, providing a solid data foundation for subsequent intelligent test paper compilation, and ensuring in-depth understanding and effective utilization of various questions.

[0100] Accurate question similarity calculation: Based on the cosine similarity calculation method of the keyword set and the image feature vector set, as well as the weighted summation method for case analysis multi-modal questions, the accurate quantification of the similarity of questions of different types is realized. This helps to avoid a large number of similar questions in the test paper compilation, ensures the richness and diversity of the test paper question content, and enables the exam to comprehensively examine the examinees' knowledge mastery.

[0101] Intelligent and flexible test paper compilation strategy: Multi-dimensional screening is carried out according to parameters such as the preset question type structure, knowledge point coverage, and difficulty level for the exam purpose. Combining the similarity threshold judgment and the optimization and replacement mechanism can not only quickly generate a test paper that meets the exam requirements, but also ensure that the difficulty distribution of the test paper questions is reasonable, the knowledge point coverage is comprehensive, and the similarity between questions is within a reasonable range. At the same time, the replacement object is determined according to the historical usage times of the questions, further improving the rationality and scientificity of the test paper compilation and the utilization efficiency of the question bank resources. Brief Description of the Drawings

[0102] The present invention will be further described below with reference to the accompanying drawings.

[0103] Figure 1 It is a flowchart of an intelligent test paper compilation method for medical theory exams based on artificial intelligence technology according to the present invention;

[0104] Figure 2 It is a flowchart of intelligent test paper compilation in an intelligent test paper compilation method for medical theory exams based on artificial intelligence technology according to the present invention. Detailed Embodiments

[0105] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0106] As Embodiment 1 of the present invention:

[0107] Please refer to Figure 1 and Figure 2 As shown, the present invention is an intelligent test paper compilation method for medical theory exams based on artificial intelligence technology, including the following steps:

[0108] The first step, preprocessing of question information:

[0109] For text-based questions in medical exam questions;

[0110] Directly extract the text content and represent it with the text set W = {W(i1, j1)};

[0111] Among them, i1 = 1, 2, ……, n1, where n1 represents the number of text-based questions, j1 = 1, 2, ……, m1, and m1 represents the number of characters in the i1-th text-based question; and W(i1, j1) represents the j1-th character in the i1-th text-based question.

[0112] For W = {W(i1, j1)}, text features are obtained using keyword extraction technology. The specific steps are as follows:

[0113] Step 1.2.1, Word Segmentation:

[0114] Using the word segmentation technology in natural language processing, the text is segmented into independent words.

[0115] Step 1.2.2, Calculate Word Frequency:

[0116] Count the number of times each word appears in each question containing text content, denoted as C(i, j).

[0117] Among them, i ∈ {i1, i3}, j ∈ {j1, j a};

[0118] Through , calculate the frequency of the j-th word in the i-th document, TF(i, j).

[0119] Among them, m ∈ {m1, m a};

[0120] Step 1.2.3, Calculate Inverse Document Frequency:

[0121] Then, select each word one by one, count the number of documents containing the corresponding word in the questions with text content, and denote it as Dj.

[0122] Next, through , calculate the inverse document frequency of each word, IDF(j).

[0123] Step 1.2.4, Determine Keywords:

[0124] After that, through , calculate the term frequency-inverse document frequency of each word, TF-IDF(i, j).

[0125] Then, select the words with TF-IDF(i, j) higher than the corresponding preset TF-IDF threshold to form a keyword set K, which is the feature vector of the text.

[0126] Second step, Calculate Question Similarity:

[0127] Calculate Text Similarity:

[0128] Cosine similarity calculation method based on a set of keywords, the steps are as follows:

[0129] Select two text titles and extract their corresponding sets of keywords Ka = {Ka(g1)} and Kb = {Ka(g2)} and convert them into binary vectors;

[0130] Where g2 = 1, 2, …… v1, v1 represents the number of all keywords in the keyword set Ka, g2 = 1, 2, …… v2; v2 represents the number of all keywords in the keyword set Kb;

[0131] Where the vector length of the binary vector is the number r of all non-repeating keywords in the two texts, the presence of a keyword is recorded as 1, and the absence is recorded as 0;

[0132] Obtain the binary vectors: Va = [Va(f)] and Vb = [Vb(f)];

[0133] Where f = 1, 2, …… r;

[0134] Illustrative example: Suppose the keyword sets of two texts are:

[0135] K1 = {"heart disease", "treatment", "symptoms"};

[0136] K2 = {"diabetes", "treatment", "complications"};

[0137] All non-repeating keywords in texts K1 and K2 are: {"heart disease", "treatment", "symptoms", "diabetes", "complications"}, that is, the vector length is r = 5;

[0138] For the binary vector of text K1:

[0139] "heart disease" ∈ K1, "treatment" ∈ K1, "symptoms" ∈ K1, "diabetes" ∉ K1, "complications" ∉ K1;

[0140] Then its binary vector is V1 = [1, 1, 1, 0, 0];

[0141] "heart disease" ∉ K1, "treatment" ∈ K1, "symptoms" ∉ K1, "diabetes" ∈ K1, "complications" ∈ K1;

[0142] Then its binary vector is V1 = [0, 1, 0, 0, 0];

[0143] Through: , calculate the text similarity SK between the two text titles;

[0144] Where the larger the value of SK, the higher the text similarity, and the value range is [0, 1];

[0145] Step 3, intelligent test paper compilation:

[0146] Step R1, extract test paper compilation requirements:

[0147] Extract the following parameters preset according to the test purpose:

[0148] Question type structure: the number of text-based questions Nx;

[0149] Knowledge point coverage: the knowledge point set S = {S(q)}, and each knowledge point S(q) has at least e(q) questions covering it, where e(q) is a preset value;

[0150] Difficulty level: The questions are divided into three levels: low, medium, and high, and the number of questions at each level is N1, N2, N3 in sequence, satisfying N = N1 + N2 + N3;

[0151] Step R2, question screening:

[0152] Step R2.1, question type screening:

[0153] Select text-based questions that meet the quantity requirements from the pre-established question bank respectively, and form a question set Tx, and it satisfies: ∣Tx∣ = Nx;

[0154] Among them, ∣Tx∣ refers to the quantity corresponding to the text-based questions in the question set;

[0155] Step R2.2, knowledge point screening:

[0156] For each knowledge point S(q), select the questions related to this knowledge point from Tx, and at the same time satisfy that the knowledge point S(q) has at least e(q) questions covering it;

[0157] Step R2.3, difficulty screening:

[0158] According to the preset difficulty level of the questions, select the questions that meet the difficulty quantity requirements from the screened questions, and form question sets T1, T2, T3, and it satisfies:

[0159] ∣T1∣ = N1, ∣T2∣ = N2, ∣T3∣ = N3;

[0160] Among them, ∣T1∣, ∣T2∣, ∣T3∣ refer to the quantities corresponding to the questions of each difficulty level in the question set;

[0161] Step R3, question combination optimization:

[0162] Step R3.1, similarity extraction:

[0163] For the question sets of the same question type, extract the similarity SK between two questions;

[0164] Step R3.2, Threshold Judgment:

[0165] Extract the preset similarity threshold SKy. Between two questions in the question set of the same question type, when their similarity SK > similarity threshold SKy, they are determined as highly similar question pairs;

[0166] Then randomly select one question from the remaining questions of the original screening conditions, that is, questions that meet the requirements of question type structure, knowledge point coverage, and difficulty level corresponding to the test paper compilation but have not been selected, and replace any one question in the highly similar question pair;

[0167] Next, recalculate the similarity between the replaced question and other questions until the similarity SK of all question pairs is less than or equal to the similarity threshold SKy, and obtain the test paper compilation question set;

[0168] Example 1 proposes an intelligent test paper compilation method for medical theory examinations based on artificial intelligence technology. In the preprocessing link of question information, text features are extracted through steps such as word segmentation, calculation of word frequency, and inverse document frequency, which can accurately mine the key information of text-based questions and lay a solid foundation for subsequent calculation of question similarity. In the calculation of question similarity, based on the cosine similarity calculation method of keyword sets, the keyword sets are converted into binary vectors for calculation, which can intuitively and efficiently measure the similarity between text-based questions. During the intelligent test paper compilation process, question type screening, knowledge point screening, and difficulty screening are carried out in sequence to ensure that the test paper compilation meets various requirements. Finally, the question combination is optimized through similarity extraction and threshold judgment, effectively reducing the question similarity, improving the quality of the test paper, making the test paper more reasonable in terms of question type structure, knowledge point coverage, and difficulty distribution, and meeting the professional and scientific requirements of medical examinations.

[0169] As Example 2 of the present invention:

[0170] Please refer to Figure 1 and Figure 2 As shown, when the present application is specifically implemented, compared with Example 1, the technical solution of this example is only different from that of Example 1 in that in this example:

[0171] One of the questions replaced in the highly similar question pair is determined by its historical usage times in the question bank. Specifically, the question with the largest historical usage times value is replaced;

[0172] In Example 2 compared with Example 1, when dealing with the replacement problem of highly similar question pairs, instead of randomly selecting questions for replacement, the replacement object is determined based on the historical usage times of the questions in the question bank, and the question with the largest historical usage times value is replaced. This improvement makes the test paper generation process more targeted and reasonable. By preferentially replacing the questions with high usage frequencies, it helps to improve the utilization rate of the questions in the question bank, avoid repeated use of high-frequency questions, increase the freshness and diversity of the questions in the test paper, and at the same time can more fairly examine the candidates' mastery of different questions, further optimize the test paper generation effect, and improve the quality and practicality of the test paper.

[0173] As Example 3 of the present invention:

[0174] Please refer to Figure 1 and Figure 2 As shown, when the present application is specifically implemented, compared with Example 1 and Example 2, the technical solution of this example lies in combining the solutions of the above Example 1 and Example 2 for implementation. The difference between the technical solution of this example and Example 1 and Example 2 is only in this example;

[0175] Step 1. Preprocessing of question information:

[0176] For the image-based questions in medical examination questions, in this example, the image-based questions are such as pathological section image questions;

[0177] Directly extract the original image data and represent it with the image set X = {X(i2, j2)};

[0178] Among them, i2 = 1, 2,... n2, n2 represents the number of image-based questions, j2 = 1, 2,... m2, m2 represents the number of images in the i-th image-based question; and X(i2, j2) represents the j2-th image in the i2-th image-based question;

[0179] For X = {X(i2, j2)} and XW = {X(i3, j b )}, the local feature descriptor method is used to extract image features, and the specific steps are as follows:

[0180] Step 1.3.1. Image segmentation:

[0181] Select an image and divide the image into s local regions;

[0182] Step 1.3.2. Color feature extraction:

[0183] Convert the image from the RGB color space to the HSV color space, and then calculate the color mean and variance of each local region:

[0184] Specifically include:

[0185] The mean value HP(k) and variance HF(k) of hue H;

[0186] The mean value SP(k) and variance SF(k) of saturation S;

[0187] The mean value VP(k) and variance VF(k) of brightness V;

[0188] where k = 1, 2, …… s;

[0189] In this embodiment, the calculation of the mean value and variance of hue H, saturation S, and brightness V belongs to the basic technology in the field of image processing. Its mathematical principle is prior art. Each pixel point contains H / S / V values, which are the inherent attributes of the image in the HSV color space and can be obtained without additional technological innovation. It belongs to the common knowledge of those skilled in the art, so no detailed derivation is made here;

[0190] Step 1.3.3, Texture feature extraction:

[0191] Use the gray-level co-occurrence matrix to calculate the texture parameters of each local area in the image;

[0192] Specifically include:

[0193] Energy E(k), entropy H(k), contrast C(k), correlation R(k);

[0194] Step 1.3.4, Construct feature vectors:

[0195] Combine the color features and texture features extracted from the local area to form the feature vector of the local area:

[0196] That is, F(k) = [HP(k), HF(k), SP(k), SF(k), VP(k), VF(k), E(k), H(k), C(k), R(k)];

[0197] Then, combine the feature vectors corresponding to all local areas included in the image to form the image feature vector set, that is, F = [F(k) = F1, F2, …… Fs];

[0198] In this embodiment, texture feature extraction belongs to the mature technology in the fields of image processing and computer vision. Its core method: the gray-level co-occurrence matrix and the calculation of its derived texture parameters belong to prior art. Its mathematical principle, parameter definition, and application scenario have been fully disclosed, so no detailed derivation is made here;

[0199] Second step, Topic similarity calculation:

[0200] Image similarity calculation:

[0201] Based on the cosine similarity calculation method of the image feature vector set, the steps are as follows:

[0202] Select two image - type questions, and extract the corresponding keyword sets Fa = {Fa(k)} and Fb = {Fb(k)} respectively, and convert them into binary vectors;

[0203] By: , calculate the image similarity SF between the two image - type questions;

[0204] Among them, the larger the SF value, the higher the image similarity, and the value range is [0, 1];

[0205] The third step, intelligent test paper generation:

[0206] Step R1, extract the test paper requirements:

[0207] Extract the following parameters preset according to the test purpose:

[0208] Question type structure: the number of image - type questions Ny;

[0209] Knowledge point coverage: the knowledge point set S = {S(q)}, and each knowledge point S(q) has at least e(q) questions covering it, where e(q) is a preset value;

[0210] Difficulty level: Divide the questions into three levels: low, medium, and high, and the number of questions at each level is N1, N2, N3 in turn, satisfying N = N1 + N2 + N3;

[0211] Step R2, question screening:

[0212] Step R2.1, question type screening:

[0213] Select the image - type questions that meet the quantity requirements from the pre - established question bank respectively, and form a question set Ty, and it satisfies: ∣Ty∣ = Ny;

[0214] Among them, ∣Ty∣ refers to the quantity corresponding to the image - type questions in the question set;

[0215] Step R2.2, knowledge point screening:

[0216] For each knowledge point S(q), select the questions involving this knowledge point from Ty, and at the same time satisfy that the knowledge point S(q) has at least e(q) questions covering it;

[0217] Step R2.3, difficulty screening:

[0218] According to the preset difficulty level of the questions, select the questions that meet the difficulty quantity requirements from the screened questions, and form question sets T1, T2, T3, and it satisfies:

[0219] ∣T1∣ = N1, ∣T2∣ = N2, ∣T3∣ = N3;

[0220] Among them, ∣T1∣, ∣T2∣, and ∣T3∣ represent the respective quantities corresponding to the questions of each difficulty level in the question set.

[0221] Step R3, question combination optimization:

[0222] Step R3.1, similarity extraction:

[0223] For the question set of the same question type, extract the similarity SF between every two questions.

[0224] Step R3.2, threshold judgment:

[0225] Extract the preset similarity threshold SFy.

[0226] Between two questions in the question set of the same question type, when their similarity SF > similarity threshold SFy, they are determined as a highly similar question pair.

[0227] Then randomly select one question from the remaining questions of the original screening conditions, that is, a question that meets the requirements of question type structure, knowledge point coverage, and difficulty level correspondence for test paper compilation but has not been selected, and replace any one question in the highly similar question pair.

[0228] Then recalculate the similarity between the replaced question and other questions until the similarity SF of all question pairs is less than or equal to the similarity threshold SFy, and obtain the test paper question set.

[0229] Example 3 combines the solutions of Example 1 and Example 2. It can not only effectively perform information preprocessing, similarity calculation, and intelligent test paper compilation for text questions, but also for image questions (such as pathological section image questions), extract image features by means of image block division, extraction of color features and texture features, etc., and then measure the similarity of image questions based on the cosine similarity calculation method of the image feature vector set, making the test paper compilation method applicable to image questions. During intelligent test paper compilation, the question type, knowledge points, and difficulty are also screened, and the question combination is optimized. In addition, when dealing with the replacement of highly similar question pairs, the method of Example 2 is adopted, and the replacement question is determined according to the historical usage times. This example expands the applicable scope of the test paper compilation method, enabling it to process both text questions and image questions simultaneously, and is more scientific in question replacement, further improving the comprehensiveness and rationality of test paper compilation, and meeting the test paper compilation requirements for different types of questions in medical examinations.

[0230] As Example 4 of the present invention:

[0231] Please refer to Figure 1 and Figure 2 As shown, when the present application is specifically implemented, compared with Example 1, Example 2, and Example 3, the difference between this example and Example 1, Example 2, and Example 3 is only that in this example, it further includes:

[0232] Step 1: Preprocessing of topic information:

[0233] For multimodal questions on case analysis in medical examinations;

[0234] In this embodiment, case analysis multimodal questions are such as questions on text and image fusion;

[0235] For case analysis multimodal questions: separate the text part and the image part, and pass the text part through WX={WX(i3,j a )}, the image part is represented by XW={X(i3,j b )}express;

[0236] Among them, i3=1, 2, ... n3, n3 represents the number of case analysis multimodal questions, j a =1, 2, ...m a , m a represents the number of characters in the i3th case analysis multimodal question; j b =1, 2, ...m b , m b represents the number of images in the i3th case analysis multimodal question; and X(i3,j a ) represents the jth case analysis multimodal question in the i3th case analysis a characters, X(i3,j b ) represents the jth case analysis multimodal question in the i3th case analysis b images;

[0237] For the text and image parts of the case analysis multimodal questions, the corresponding keyword set is extracted according to the text feature extraction method in Example 1, and the corresponding image feature vector set is extracted according to the image feature extraction method in Example 2;

[0238] Step 2: Calculate the similarity of the title:

[0239] Similarity calculation for multimodal questions related to case analysis:

[0240] Based on the text and image similarity calculation method, a weighted summation method is used. The steps are as follows:

[0241] Select two case analysis multimodal questions, calculate the similarity SK0 of the corresponding text parts of the two using the text similarity calculation method, and at the same time, calculate the similarity SF0 of the corresponding image parts of the two using the image similarity calculation method;

[0242] pass: , calculate the overall similarity SZ between two case analysis multimodal questions;

[0243] Wherein, α K and α F are corresponding preset weight coefficients, and α K +α F = 1;

[0244] In this embodiment, the weight coefficients of the text and the image can be adjusted according to actual needs;

[0245] Step 3: Intelligent test paper compilation:

[0246] Step R1: Extract test paper compilation requirements:

[0247] Extract the following parameters preset according to the test purpose:

[0248] Question type structure: The number of case analysis questions Nz;

[0249] Knowledge point coverage: The knowledge point set S = {S(q)}, and each knowledge point S(q) has at least e(q) questions covering it, where e(q) is a preset value;

[0250] Difficulty level: The questions are divided into three levels: low, medium, and high, and the number of questions at each level is N1, N2, and N3 in sequence, satisfying N = N1 + N2 + N3;

[0251] Step R2: Question screening:

[0252] Step R2.1: Question type screening:

[0253] Select case analysis questions that meet the quantity requirements from the pre-established question bank respectively, and form a question set Tz, and it satisfies: |Tz| = Nz;

[0254] Among them, |Tz| refers to the quantities corresponding to text-based, image-based, and case analysis questions in the question set;

[0255] Step R2.2: Knowledge point screening:

[0256] For each knowledge point S(q), select the questions involving this knowledge point from Tz, and at the same time satisfy that the knowledge point S(q) has at least e(q) questions covering it;

[0257] Step R2.3: Difficulty screening:

[0258] According to the preset difficulty level of the questions, select the questions that meet the difficulty quantity requirements from the screened questions, and form question sets T1, T2, and T3, and it satisfies:

[0259] |T1| = N1, |T2| = N2, |T3| = N3;

[0260] Among them, ∣T1∣, ∣T2∣, and ∣T3∣ represent the respective quantities corresponding to the questions of each difficulty level in the question set.

[0261] Step R3, question combination optimization:

[0262] Step R3.1, similarity extraction:

[0263] For the question set of the same question type, extract the similarity SZ between every two questions.

[0264] Step R3.2, threshold judgment:

[0265] Extract the preset similarity threshold SZy.

[0266] Between two questions in the question set of the same question type, when their similarity SZ > similarity threshold SZy, they are determined as a highly similar question pair.

[0267] Then randomly select one question from the remaining questions of the original screening conditions, that is, a question that meets the requirements of question type structure, knowledge point coverage, and difficulty level correspondence for test paper compilation but has not been selected, and replace any one question in the highly similar question pair.

[0268] Then recalculate the similarity between the replaced question and other questions until the similarity SZ of all question pairs is less than or equal to the similarity threshold SZy, and obtain the question set for test paper compilation.

[0269] In Example 4 for case analysis multi-modal questions (such as text and image fusion questions), during the preprocessing of question information, the text part and the image part are separated, and the text features and image features are extracted according to the methods of Example 1 and Example 2 respectively. In the calculation of question similarity, the weighted summation method is used to comprehensively obtain the overall similarity of the text and the image, which can more comprehensively and accurately measure the similarity degree of multi-modal questions. During the intelligent test paper compilation process, for case analysis questions, question type, knowledge point, and difficulty screening are performed, and the question combination is optimized through similarity extraction and threshold judgment. This example solves the information processing and similarity calculation problems of multi-modal questions during the test paper compilation process, enables the test paper compilation method to be applicable to more complex question types, enriches the question types of test paper compilation, improves the examination ability of the test paper for medical practical application scenarios, and makes the examination better reflect the comprehensive application ability of candidates' knowledge.

[0270] As Example 5 of the present invention:

[0271] Please refer to Figure 1 and Figure 2 As shown, when the present application is specifically implemented, compared with Example 1, Example 2, Example 3, and Example 4, the technical solution of this example lies in combining and implementing the solutions of the above Example 1, Example 2, Example 3, and Example 4.

[0272] Example 5 The solutions of Examples 1, 2, 3, and 4 are combined and implemented, which has the ability to handle various types of questions, such as text-based, image-based, and case analysis-based multi-modal questions. In terms of preprocessing the question information, corresponding feature extraction methods can be adopted for different types of questions; when calculating the question similarity, appropriate calculation methods can be selected according to the question type, such as text similarity calculation, image similarity calculation, overall similarity calculation of multi-modal questions, etc.; in the process of intelligent test paper compilation, on the basis of meeting the requirements of question type structure, knowledge point coverage, difficulty level, etc., for the replacement of highly similar question pairs, it can be determined according to the historical usage times, making the test paper compilation process more scientific and reasonable. This example combines the advantages of each example, greatly expands the application scope of the test paper compilation method, improves the flexibility and comprehensiveness of test paper compilation, and can meet the diverse test paper compilation needs in medical examinations, generating high-quality test papers with more reasonable structures, richer contents, and more comprehensive examinations.

[0273] It should be stated that: all user data collected in this application is collected with the consent and authorization of the users, and the uses of the user data are legal and compliant, and the use and processing of the user data comply with the relevant laws, regulations and standards of the relevant regions.

[0274] The above formulas are all dimensionless and take their numerical values for calculation. The formula is a formula obtained by collecting a large amount of data and performing software simulation to approximate the real situation. The preset parameters and threshold selection in the formula are set by those skilled in the art according to the actual situation.

[0275] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or replacements, which should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claimed rights.

Claims

1. An intelligent test paper generation method for medical theory examinations based on artificial intelligence technology, characterized in that, It includes the following steps: The first step, preprocessing of question information: Decompose the questions by question type into multi-modal questions of text type, image type, and case analysis type, and extract their features. The feature extraction includes: Text feature extraction: Segment the text part of text type and case analysis type questions, calculate word frequencies and inverse document frequencies, and determine the keyword set; Image feature extraction: Divide the image part of image type and case analysis type questions into blocks, extract color and texture features, and construct an image feature vector set; The second step, calculation of question similarity: Based on the keyword set and the feature vector set, calculate text similarity, image similarity, and overall multi-modal question similarity respectively; The third step, intelligent test paper compilation: Screen questions according to the preset question type structure, knowledge point coverage, and difficulty level requirements, and optimize the combination to obtain a set of test paper questions.

2. The intelligent test paper generation method for medical theory examinations based on artificial intelligence technology according to claim 1, characterized in that, For text type questions: Directly extract the text content and represent it with a text set W = {W(i1, j1)}; Among them, i1 = 1, 2,... n1, n1 represents the number of text type questions, j1 = 1, 2,... m1, m1 represents the number of words in the i1-th text type question; and W(i1, j1) represents the j1-th word in the i1-th text type question; For image type questions: Directly extract the original image data and represent it with an image set X = {X(i2, j2)}; Among them, i2 = 1, 2,... n2, n2 represents the number of image type questions, j2 = 1, 2,... m2, m2 represents the number of images in the i-th image type question; and X(i2, j2) represents the j2-th image in the i2-th image type question; For case analysis multi-modal questions: Separate the text part and the image part, and represent the text part by WX = {WX(i3, j a ), and represent the image part by XW = {X(i3, j b )}; where \(i_3 = 1, 2, \cdots, n_3\), \(n_3\) represents the number of case analysis type multi-modal questions, and \(j\) a = 1, 2, \cdots, m a , \(m\) a represents the number of words in the \(i_3\)-th case analysis type multi-modal question; \(j\) b = 1, 2, \cdots, m b , \(m\) b represents the number of images in the \(i_3\)-th case analysis type multi-modal question; and \(X(i_3, j\) a ) represents the \(j\)-th a word in the \(i_3\)-th case analysis type multi-modal question, and \(X(i_3, j\) b ) represents the \(j\)-th b image in the \(i_3\)-th case analysis type multi-modal question.

3. The intelligent test paper generation method for medical theory examinations based on artificial intelligence technology according to claim 2, characterized in that, The text feature extraction method is as follows: For W = {W(i1, j1)} and WX = {WX(i3, j a )}, the keyword extraction technology is adopted to obtain the text features, and the specific steps are as follows: Use the word segmentation technology in natural language processing to segment the text into independent words; Count the number of occurrences C(i, j) of each word in each question containing text content; where i ∈ {i1, i3} and j ∈ {j1, j a}; Through , the frequency TF(i, j) of the j-th word appearing in the i-th document is calculated; where, m ∈ {m1, m a}; then each word is selected one by one, and the number of documents containing the corresponding word in the title with text content is counted and denoted as Dj; Next, through , the inverse document frequency IDF(j) of each word is calculated; then through , the term frequency-inverse document frequency TF-IDF(i, j) of each word is calculated; subsequently, the words with TF-IDF(i, j) higher than the corresponding preset TF-IDF threshold are selected to form the keyword set K.

4. The intelligent test paper generation method for medical theory examinations based on artificial intelligence technology according to claim 3, characterized in that, The image feature extraction method is as follows: For X = {X(i2, j2)} and XW = {X(i3, j b )}, the local feature descriptor method is used to extract image features, and the specific steps are as follows: Select an image and divide the image into s local regions; Convert the image from the RGB color space to the HSV color space, and then calculate the color mean and variance of each local region: Use the gray-level co-occurrence matrix to calculate the texture parameters of each local region in the image; Combine the color features and texture features extracted in the local region to form the feature vector of the local region: Then combine the feature vectors corresponding to all local regions included in the image to form an image feature vector set.

5. The intelligent test paper generation method for medical theory examinations based on artificial intelligence technology according to claim 4, wherein The color mean and variance of each local region include: the mean HP(k) and variance HF(k) of hue H; the mean SP(k) and variance SF(k) of saturation S; the mean VP(k) and variance VF(k) of brightness V; where k = 1, 2,... s; The texture parameters of each local region include: energy E(k), entropy H(k), contrast C(k), correlation R(k); The feature vector of the local region is F(k) = [HP(k), HF(k), SP(k), SF(k), VP(k), VF(k), E(k), H(k), C(k), R(k)]; The image feature vector set is F = [F(k) = F1, F2,... Fs].

6. The intelligent test paper generation method for medical theory examinations based on artificial intelligence technology according to claim 4, characterized in that, The text similarity calculation method is as follows: Based on the cosine similarity calculation method of the keyword set, the steps are as follows: Select two text-based questions, and extract their corresponding keyword sets Ka = {Ka(g1)} and Kb = {Ka(g2)}, and convert them into binary vectors; Among them, g2 = 1, 2, …… v1, where v1 represents the number of all keywords in the keyword set Ka, and g2 = 1, 2, …… v2; v2 represents the number of all keywords in the keyword set Kb; Among them, the vector length of the binary vector is the number r of all non-repeating keywords in the two texts. If a keyword appears, it is recorded as 1, and if it does not appear, it is recorded as 0; Obtain binary vectors: Va = [Va(f)] and Vb = [Vb(f)]; Among them, f = 1, 2, …… r; Adopted by: , the text similarity SK between two text-based questions is calculated.

7. A method for intelligent test paper generation for medical theory examinations based on artificial intelligence technology according to claim 6, characterized in that, The calculation method of image similarity is as follows: Based on the cosine similarity calculation method of the image feature vector set, the steps are as follows: Select two image-based questions, and extract their corresponding keyword sets Fa = {Fa(k)} and Fb = {Fb(k)}, and convert them into binary vectors; Adopted by: , the image similarity SF between two image-based questions is calculated.

8. A method for intelligently generating test papers for medical theory examinations based on artificial intelligence technology according to claim 7, characterized in that, The calculation method of the similarity of case analysis multi-modal questions is as follows: Based on the text and image similarity calculation methods, adopt the weighted summation method, and the steps are as follows: Select two case analysis multi-modal questions, calculate the similarity SK0 of the corresponding text parts of the two according to the text similarity calculation method, and at the same time, calculate the similarity SF0 of the corresponding image parts of the two according to the image similarity calculation method; Adopted by: , the overall similarity SZ between two case analysis multi-modal questions is calculated; where α K and α F are corresponding preset weight coefficients, and α K + α F = 1.

9. A method for intelligently generating test papers for medical theory examinations based on artificial intelligence technology according to claim 8, characterized in that, The intelligent test paper generation method is as follows: Step R1: Extract the test paper generation requirements: Extract the following parameters preset according to the test purpose: Question type structure: The total number of questions is N, and the question type distribution is the number Nx of text-based questions, the number Ny of image-based questions, and the number Nz of case analysis questions, and it satisfies N = Nx + Ny + Nz; Knowledge point coverage: The knowledge point set S = {S(q)}, and each knowledge point S(q) has at least e(q) questions covering it, where e(q) is a preset value; Difficulty level: The questions are divided into three levels: low, medium, and high, and the number of questions at each level is N1, N2, N3 in turn, satisfying N = N1 + N2 + N3; Step R2: Question screening: Question type screening: Select text-based, image-based, and case analysis questions that meet the quantity requirements from the pre-established question bank respectively, and form question sets Tx, Ty, Tz, and it satisfies: ∣Tx∣ = Nx, ∣Ty∣ = Ny, ∣Tz∣ = Nz, where ∣Tx∣, ∣Ty∣, ∣Tz∣ represent the quantities corresponding to text-based, image-based, and case analysis questions in the question set respectively; Knowledge point screening: For each knowledge point S(q), select the questions involving this knowledge point from Tx, Ty, Tz, and at the same time satisfy that the knowledge point S(q) has at least e(q) questions covering it; Difficulty screening: According to the preset difficulty level of the questions, select the questions that meet the difficulty quantity requirements from the screened questions, and form question sets T1, T2, T3, and it satisfies: ∣T1∣ = N1, ∣T2∣ = N2, ∣T3∣ = N3, where ∣T1∣, ∣T2∣, ∣T3∣ represent the quantities corresponding to questions of each difficulty level in the question set respectively; Step R3: Question combination optimization: Similarity extraction: For a set of questions of the same question type, extract the similarity S0 between every two questions, where S0 ∈ {SK, SF, SZ}. Threshold judgment: Extract the pre-set similarity threshold S0y, where S0y ∈ {SKy, SFy, SZy}; SKy represents the text threshold, SFy represents the image threshold, and SZy represents the multi-modal threshold. Between two questions in a set of questions of the same question type, when their similarity S0 is greater than the similarity threshold S0y, they are determined to be a highly similar question pair. Then randomly select one question from the remaining questions of the original screening conditions, that is, a question that meets the requirements of question type structure, knowledge point coverage, and difficulty level correspondence but has not been selected, and replace one question in the highly similar question pair. Then recalculate the similarity between the replaced question and other questions until the similarity S0 of all question pairs is less than or equal to the similarity threshold S0y, and obtain the set of questions for the test paper.

10. A method for intelligent test paper generation for medical theory examinations based on artificial intelligence technology according to claim 9, characterized in that, Among them, One question in the highly similar question pair to be replaced is determined by its historical usage times in the question bank. Specifically, the question with the largest historical usage times value is replaced.

Images