Student ability diagnosis method and related equipment and storage media

Abstract

The present application discloses a method for diagnosing student abilities and related equipment and storage media, wherein the method for diagnosing student abilities comprises: a computing device inputs a test answer result; the computing device decomposes the test answer result into steps to obtain a plurality of answer steps; the computing device uses a network model to analyze the answer steps to determine a correct / wrong type label; the computing device forms an ability diagnosis result based on the correct / wrong type label. The above scheme can intelligently diagnose student abilities.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a student ability diagnosis method and related equipment and storage media. Background Art

[0002] In order to uphold the teaching philosophy of teaching students in accordance with their aptitude, we hope to judge students' abilities and truly understand the strengths or weaknesses of each student, so as to cultivate students' abilities in a more targeted manner. At present, teachers can evaluate students' abilities based on their own experience, classroom performance and test answers. This reliance on manual diagnosis of students' abilities is basically entirely dependent on the teacher's subjective evaluation, which obviously consumes a lot of manpower.

[0003] In view of this, how to intelligently diagnose students' abilities has become an urgent problem to be solved. Summary of the invention

[0004] The main technical problem solved by this application is to provide a student ability diagnosis method and related equipment and storage medium, which can intelligently diagnose student abilities.

[0005] In order to solve the above-mentioned technical problems, the first aspect of the present application provides a student ability diagnosis method, including: a computing device inputs a test answer result; the computing device decomposes the test answer result into steps to obtain a plurality of answer steps; the computing device uses a network model to analyze the answer steps to determine a correct / incorrect type label; the computing device forms an ability diagnosis result based on the correct / incorrect type label.

[0006] In order to solve the above-mentioned technical problems, the second aspect of the present application provides a computing device, including: an input module for inputting test answer results; a disassembly module for step-by-step decomposition of the test answer results to obtain a number of answer steps; a determination module for analyzing the answer steps using a network model to determine correct / incorrect type labels; and a result formation module for forming an ability diagnosis result based on the correct / incorrect type labels.

[0007] In order to solve the above technical problems, the third aspect of the present application provides an electronic device, including a memory and a processor coupled to each other, wherein the memory stores program instructions, and the processor is used to execute the program instructions to implement the student ability diagnosis method described in the first aspect above.

[0008] In order to solve the above technical problems, the fourth aspect of the present application provides a computer-readable storage medium storing program instructions that can be executed by a processor, and the program instructions are used to implement the student ability diagnosis method described in the first aspect above.

[0009] In the above scheme, after the computing device inputs the test answer results, the test answer results are broken down into steps to obtain several answer steps, and then the answer steps are analyzed using a network model to determine the correct / incorrect type labels. Finally, based on the correct / incorrect type labels, an ability diagnosis result is formed. The entire process is completed by the computing device, thereby being able to intelligently diagnose students' abilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a flowchart of the first embodiment of the student ability diagnosis method of the present application;

[0011] Figure 2 It is a flow chart of the second embodiment of the student ability diagnosis method of the present application;

[0012] Figure 3 It is a flowchart of the third embodiment of the student ability diagnosis method of the present application;

[0013] Figure 4 is a flowchart of the fourth embodiment of the student ability diagnosis method of the present application;

[0014] Figure 5 is a flowchart of the fifth embodiment of the student ability diagnosis method of the present application;

[0015] Figure 6 It is a schematic diagram of the framework of an embodiment of a computing device of the present application;

[0016] Figure 7 It is a schematic diagram of the framework of an embodiment of the electronic device of the present application;

[0017] Figure 8 It is a schematic diagram of a framework of an embodiment of a computer-readable storage medium of the present application. DETAILED DESCRIPTION

[0018] The scheme of the embodiment of the present application is described in detail below in conjunction with the drawings of the specification.

[0019] In the following description, for the purpose of explanation rather than limitation, specific details such as specific system structures, interfaces, and technologies are provided to facilitate a thorough understanding of the present application.

[0020] The terms "system" and "network" are often used interchangeably in this article. The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship. In addition, "many" in this article means two or more than two.

[0021] See also Figure 1 , Figure 1 This is a flow chart of the first embodiment of the student ability diagnosis method of the present application. Specifically, it may include the following steps:

[0022] Step S11: The computing device inputs the test answer results.

[0023] The test answer result can be a test answer result obtained in real time, or a historical test answer result. The test answer result can be generated on a computing device, can be sent by an external device received by the computing device, or can be obtained from a test database. The test answer result can be of a target student, or of a student associated with the target student, wherein the target student is the object of the student ability diagnosis, and the students associated with the target student include the associated students and comparison students mentioned later. The test answer result can be one or more, and can come from the same student or multiple students. The test questions and their test answer results and other data in this article can be from various subjects, for example, mathematics. Therefore, the sources of obtaining the test answer results can be varied, and are not specifically limited here. The computing device can be a device that can implement the student ability diagnosis method, including various devices such as computers, mobile phones, and tablets, which are not specifically limited here.

[0024] Step S12: The computing device decomposes the test answer result into steps to obtain a plurality of answer steps.

[0025] The test answer result includes the answer formed by the student after answering the test question, and the test answer result may include several answer steps. In the test answer result, it may appear that the previous answer steps are correct, but errors occur in the key answer step or the last answer step. Therefore, through in-depth analysis of the student's answer process, especially the analysis of the answer steps, it can be concluded that the student's quality is weak in computing ability, lack of data analysis dimensions, incorrect use of conditions, and incorrect derivation. The computing device can decompose the test answer result step by step to obtain several answer steps in order to analyze the answer steps.

[0026] In a disclosed embodiment, the computing device performs step decomposition on the answer result of the test question, and when several answer steps are obtained, the answer result of the test question is divided into several answer steps using the preset symbols in the answer result of the test question as the boundary; the answer steps are segmented using the preset segmentation set to obtain several answer steps after segmentation. The preset symbols can be set according to the characteristics of the subject, and are not specifically limited here. For example, the preset symbols include but are not limited to symbols or characters such as "∵", "∴", "due to", "so", "derivable", etc. that represent the meaning of separation. The preset segmentation set can also be set according to the characteristics of the subject. For example, in the subject of mathematics, in order to prevent mathematical formulas, "isosceles triangles" and other proper nouns from being separated during segmentation, the preset segmentation set can include mathematical formulas, isosceles triangles, etc. Therefore, the test answer result can be decomposed into steps by preset symbols, and the answer steps can be segmented using the preset segmentation set, which can be more flexible in segmentation processing and close to the actual scene.

[0027] Step S13: The computing device uses the network model to analyze the answering steps and determine the correct / incorrect type label.

[0028] Since the answering steps can reflect the students' abilities, and the answering results of the test questions often correspond to the correct answering results, by comparing the answering results of the test questions with the correct answering results, it is possible to determine where the students are wrong or correct. Therefore, the computing device can use the network model to analyze the answering steps and determine the correct / wrong type labels. Of course, only the correct type labels can be determined to obtain the wrong type labels; only the wrong type labels can be determined to obtain the correct type labels; and the wrong type labels and the correct type labels can also be determined at the same time.

[0029] In a disclosed embodiment, when the computing device uses the network model to analyze the answering steps and determines the correct / incorrect type labels, the correct / incorrect type labels of the scoring points can be determined based on the answering steps, and / or the correct / incorrect type labels of the step relationships can be determined based on the relationships between different answering steps. Therefore, analysis can be performed for each answering step, or for the relationships between different answering steps, thereby increasing the depth of analysis of the test answering results.

[0030] When determining the correct / wrong type label of the scoring point based on the answering steps, the scoring point extraction model can be used to extract the scoring points of the answering steps to obtain at least one detection scoring point, and then determine whether the detection scoring point matches the standard scoring point, wherein there is a scoring point mapping relationship between the standard scoring point and the correct / wrong type label; if so, the correct / wrong type label of the detection scoring point is determined based on the scoring point mapping relationship. The scoring point extraction model can be a sequence-to-sequence model, or other model forms that can identify the scoring points of the answering steps, which are not specifically limited here. The wrong type label of the detection scoring point includes but is not limited to weak computing power, lack of data analysis dimensions, etc. The number of detection scoring points extracted from the answering steps may be one or more. By judging whether the detection scoring point matches the standard scoring point, the detection scoring point is compared with the correct standard scoring point for word segmentation, and the correct / wrong type label corresponding to the student can be output, so that the results of each correct / wrong type label can be clustered and statistically analyzed later to obtain whether the student has weak computing power, lack of data analysis dimensions, etc.

[0031] In subjects such as mathematics, since the answer results of the test questions may include the step relationships of answering steps such as conditions and derivations, the students' reasoning and modeling abilities can be judged by analyzing the relationships between different answering steps. When determining the correct / wrong type labels of the step relationships based on the relationships between different answering steps, the text structure parsing model can be used to parse the detection step relationships between different answering steps, and then determine whether the detection step relationships match the standard step relationships, wherein there is a step relationship mapping relationship between the standard step relationships and the correct / wrong type labels; if so, the correct / wrong type labels of the detection step relationships are determined based on the step relationship mapping relationship. Detection step relationships include but are not limited to conditions and derivations. Error type labels of detection step relationships include but are not limited to conditional usage errors, derivation errors, etc., and error type labels of other step relationships can be customized. The text structure parsing model can determine the relationship between the upper and lower answering steps. For example, when using the text structure parsing model to parse the detection step relationships between different answering steps, the beam search of the text structure parsing model can be used to determine the detection step relationship with the highest conditional probability. Therefore, when the paragraph structure parsing model parses the detection step relationship between different answering steps, it combines beam search to find the optimal path, find the sequence with the largest reasonable conditional probability of the corresponding answering steps, and infer the most reasonable next answering step, thereby completing the parsing of the detection step relationship.

[0032] Step S14: The computing device generates a capability diagnosis result based on the correct / incorrect type label.

[0033] After determining the correct / incorrect type label, the computing device can form an ability diagnosis result based on the correct / incorrect type label. Each test answer result may have one or more correct / incorrect type labels. When forming the ability diagnosis result, the types and / or quantities of correct / incorrect type labels can be counted to obtain an ability diagnosis result based on student literacy. For example, if there is an incorrect type label or there are more than a preset number of incorrect type labels, the difference between the default score and the preset score is used as the ability diagnosis result, or the correct / incorrect type label and the test answer result are combined to form a clustered statistical test score for the corresponding test question to obtain the ability diagnosis result.

[0034] In a public embodiment, when the computing device inputs the answer results of the test questions, the computing device inputs the answer results of several test questions so as to analyze the multiple test questions. Based on this, when the computing device forms the ability diagnosis results based on the correct / wrong type labels, it first determines the ratio of the test scores of all the test questions corresponding to the wrong type labels to the standard scores of the test questions in several test questions; then obtains the average value of the ratios of the corresponding wrong type labels in several test questions, and uses the average value as the score of the wrong type label to form the ability diagnosis results corresponding to each wrong type label. Therefore, using the correct / wrong type labels, the test scores of the test questions and the standard scores of the test questions, under each wrong type label, the average value of the ratios of all the test questions belonging to the wrong type label is calculated, and the score corresponding to each wrong type label is obtained to form the ability diagnosis results, thereby comprehensively determining the ability of the students by the test answer results of multiple test questions. When the computing device forms an ability diagnosis result based on the correct / incorrect type label, it can also count the first sum of the test scores of all test questions under the corresponding error type label and the second sum of the test standard scores, and take the ratio of the first sum to the second sum as the ability diagnosis result corresponding to the error type label.

[0035] Since there may be large accidental errors when directly diagnosing the student ability of the target student, the target student can be compared with the comparison student to assist in judging the ability diagnosis of the target student. The student ability diagnosis method of this scheme is also performed on the test answer results of several test questions of several comparison students to obtain the ability diagnosis results corresponding to each error type label of several comparison students, so as to assist in judging the ability diagnosis of the target student by using the ability diagnosis results of the comparison students. In a public embodiment, when the computing device inputs the test answer results, the computing device inputs the test answer results of several test questions of the target student and several comparison students, so that when the computing device forms the ability diagnosis results based on the correct / wrong type label, the average value of the ability diagnosis results corresponding to each error type label of several comparison students is obtained, so as to sum and average the scores of several comparison students in each ability; the difference between the ability diagnosis result of the target student and the average value of the ability diagnosis result is used as the ability diagnosis result corresponding to the error type label of the target student. Based on this, when diagnosing the student ability of the target student, the ability diagnosis result of the comparison student is included, and the difference between the ability diagnosis result of the target student and the average value of the ability diagnosis result is used as the final ability diagnosis result of the target student, which can reduce the random error caused by data sparsity.

[0036] The method for determining the comparison students can be set as needed. In an actual scenario, the ratio of the test scores of all test questions of the target student within the diagnostic time interval to the test standard score is obtained, and the average of the ratios of all test questions is used as the first initial ability score of the target student, wherein the diagnostic time interval can be one semester, one month, etc. Then, the second initial ability scores of several students within the corresponding diagnostic time interval are obtained, wherein the method for obtaining the second initial ability score is the same as the method for obtaining the first initial ability score. Then, the first initial ability score and several second initial ability scores are ranked, so that students with a preset score ratio above and below the first initial ability score are defined as students with the same score, and students with a preset ranking ratio above and below the ranking of the target student are defined as students with the same ranking, that is, students with the same score and students with the same ranking can be used together as the above-mentioned comparison students, or the intersection of students with the same score and students with the same ranking can be used as the above-mentioned comparison students. The preset score ratio and the preset ranking ratio can be customized, such as 15%, 30%, etc.

[0037] In the above scheme, after the computing device inputs the test answer results, the test answer results are broken down into steps to obtain several answer steps, and then the answer steps are analyzed using a network model to determine the correct / incorrect type labels. Finally, based on the correct / incorrect type labels, an ability diagnosis result is formed. The entire process is completed by the computing device, thereby being able to intelligently diagnose students' abilities.

[0038] In order to achieve comprehensive diagnosis of multi-dimensional student abilities, when the computing device forms an ability diagnosis result based on the correct / wrong type label, the obtained ability diagnosis result is based on the student's literacy. The computing device can also obtain and use at least one of the ability diagnosis results based on ideas and methods, the ability diagnosis results based on problem-solving methods, the ability diagnosis results based on weak knowledge points, and the ability diagnosis results based on learning attitudes to form a comprehensive ability diagnosis result. The following describes the acquisition of ability diagnosis results based on ideas and methods, ability diagnosis results based on problem-solving methods, ability diagnosis results based on weak knowledge points, and ability diagnosis results based on learning attitudes.

[0039] See also Figure 2 , Figure 2 This is a flow chart of the second embodiment of the student ability diagnosis method of the present application. Specifically, it may include the following steps:

[0040] Step S21: The computing device inputs a number of reference test questions and their corresponding reference test question scores, as well as a number of test test questions including knowledge point tags.

[0041] The reference test questions are test questions corresponding to the answer results of the above test questions. In addition, the reference test questions may also include other test questions to increase the data volume of the reference test questions and improve the accuracy of the student's ability. The computing device may obtain a number of test questions from the test question library, and the test questions are labeled with knowledge points.

[0042] Step S22: based on a plurality of reference test questions and their corresponding reference test question scores, respectively obtaining a reference test question representation of each reference test question, and based on a plurality of test test questions, respectively obtaining a test test question representation of each test test question.

[0043] In order to subsequently process various test questions, the reference test questions and the test test questions are represented as vector patterns that can be recognized by a computing device.

[0044] In a public embodiment, in order to improve the accuracy of the test question representation and make the test question representation more suitable for the actual scene, before obtaining the corresponding test question representation, the reference test question / test question can be segmented using a preset word segmentation set to obtain the reference test question / test question after segmentation. The preset word segmentation set can be set according to the characteristics of the subject. For example, in the subject of mathematics, in order to prevent mathematical formulas, "isosceles triangles" and other proper nouns from being split during word segmentation, the preset word segmentation set can include mathematical formulas, isosceles triangles, etc. By limiting the word segmentation results through the preset word segmentation set, the words in the preset word segmentation set can be prevented from being split, thereby improving the accuracy of the test question representation. For example, since there are many subject characteristics in subjects such as mathematics, for example, the meanings of mathematical letters in different expressions are completely different, therefore, through the preset word segmentation set, mathematical expressions, proper mathematical symbols, mathematical proper names, etc. can be independently identified, so that the word segmentation is more suitable for the actual application scenario.

[0045] In a disclosed embodiment, when obtaining the reference question representation of each reference question based on several reference questions and their corresponding reference question scores, the reference questions can be respectively input into the question representation acquisition model to obtain the reference question vector representation of the reference question, and the reference question scores can be respectively normalized to obtain the reference score representation of the reference question scores, and then the product of the corresponding reference question vector representation and the reference score representation is used as the reference question representation. When obtaining the test question representation of each test question based on several test questions, the test questions can be respectively input into the question representation acquisition model to obtain the test question representation of the test question. Several reference questions and their corresponding reference question scores can be processed into a two-tuple list to input into the question representation acquisition model, and the two-tuple list is, for example, [(reference question 1, reference question score 1), (reference question 2, reference question score 2)…]. During the model training process of the test question representation acquisition model, the sample test questions include various labels mentioned in this article. During the model training process, the loss function is the sum of the weighted adjustments of various labels. The weighted parameters involved use control scalars, and the specific training process also adds corresponding margins to ensure that there is distinction between classes and a certain degree of fault tolerance, so that after optimizing the minimum cross entropy, the output result of the hidden layer vector layer of the network structure before the prediction layer is used as the corresponding test question representation. Therefore, the vector representation of the intermediate process of the test question representation acquisition model is taken as the test question representation. When obtaining the reference test question representation, the product of the corresponding reference test question vector representation and the reference score representation can be used as the reference test question representation, and the result of the processing of the reference test question vector representation and the reference score representation by vector splicing can also be used as the reference test question representation. Since the reference score representation obtained by incorporating the reference test question score when obtaining the reference test question representation reflects the student's mastery of the reference test question to a certain extent, after the product of the reference test question vector representation and the reference score representation is used as the reference test question representation, the reference test question representation is the preliminary result of the personal ability assessment, forming the initial representation of the student's personal learning situation.

[0046] In a disclosed embodiment, when normalizing the reference test question scores to obtain the reference score representation of the reference test question scores, several related test question scores of several related test questions associated with the reference test question can be obtained; the mean of several related test question scores is obtained, the difference between the reference test question score and the mean is determined, and the standard deviation of the reference test question score and several related test question scores is obtained, and the quotient of the difference and the standard deviation is used as the reference score representation; or, the first difference between the reference test question score and the smallest related test question score is obtained, and the second difference between the largest related test question score and the reference test question score is obtained, and the quotient of the first difference and the second difference is used as the reference score representation. Compared with directly using the reference test question score to obtain the reference test question representation, the reference test question score has a greater impact on the vector, and after normalizing the reference test question score, not only the related test question score is used to reduce the randomness of the data, but also the normalized processing result is more conducive to the representation of the vector.

[0047] Step S23: input all reference question representations and test question representations into the score prediction model to obtain the test question score of each test question.

[0048] All reference question representations and test question representations are input into the score prediction model. The score prediction model can capture the association between similar questions based on the association between the reference question representations and the test question representations to obtain the test question score of each test question. Of course, several reference question representations and test question representations can also be selected to input into the score prediction model to obtain the test question score of each test question.

[0049] In a public embodiment, all reference question representations and test question representations are input into a score prediction model. When the test question score of each test question is obtained, the reference question representation and the test question representation can be matched to respectively determine the reference question with the greatest correlation with each test question, and the reference question score of the reference question with the greatest correlation can be used to predict the test question score of the corresponding test question. In this way, the test question score can be used to evaluate the students' existing and potential mastery of knowledge points without the students actually taking a large number of questions.

[0050] Step S24: Based on the test question scores and the corresponding knowledge point labels of the test questions, an ability diagnosis result based on weak knowledge points is formed.

[0051] After obtaining the test question scores, the ability diagnosis results based on weak knowledge points can be formed based on the test question scores and the corresponding test question knowledge point labels. For example, the total / average value of the test question scores under the corresponding knowledge point labels can be counted, and a preset number of knowledge points with the top total / average values ​​can be used as weak knowledge points.

[0052] In a public embodiment, the test questions include standard scores for the test questions. When forming an ability diagnosis result based on weak knowledge points based on the test question scores and the corresponding knowledge point labels of the test questions, the ratio of the test question scores of all test questions to the test question standard scores can be determined; the average value of the ratios of the test questions with corresponding knowledge point labels is obtained, and the average value is used as the score of the knowledge point label to form an ability diagnosis result corresponding to each knowledge point label, thereby predicting students' weak knowledge points through test question scores, without students having to actually do a large number of reference questions, and the number of reference questions can be reduced.

[0053] In a public embodiment, the reference test questions can be marked with grade weights. When the computing device inputs a number of reference test questions and their corresponding reference test question scores, a number of reference test questions and their corresponding reference test question scores are input in batches according to the grade weights of the reference test questions, so that after all reference test question representations and test question representations are input into the score prediction model and the test question score of each test question is obtained, the ability diagnosis result based on the weak knowledge point is formed based on the grade weight, the test question score and the corresponding test question knowledge point label, so that the ability diagnosis result is the result after the grade weight processing. For example, the ratio of the test question score of all test questions to the standard score of the test question is determined, and the average value of the ratio of the test questions corresponding to the knowledge point label is obtained, and the product of the grade weight and the average value is used as the score of the knowledge point label to form the ability diagnosis result corresponding to each knowledge point label. The method for determining the grade weight is not specifically limited. In an actual scenario, the grade weight is determined according to the acquisition time of the reference test questions. For example, in order to fit the memory curve logic, when determining the grade weight, the earlier the acquisition time, the smaller the grade weight. For example, the grade weight of reference test questions obtained more than one week ago is 0.15, the grade weight of reference test questions obtained more than one day but within one week is 0.35, and the grade weight of reference test questions obtained within one day is 0.5.

[0054] Through the above method, different from mapping wrong questions with knowledge points, counting the number of wrong questions or scores, and directly obtaining weak knowledge points, the statistical method is too rigid, and the number of questions affects the coverage, resulting in too much accidental factors in the diagnosis result. In this case, all reference question representations and test question representations are input into the score prediction model to obtain the test question score of each test question, and then based on the test question score and the corresponding knowledge point label of the test question, the ability diagnosis result based on weak knowledge points is formed. On the one hand, students do not need to actually do a large number of reference questions, which can reduce the number of reference questions. On the other hand, because the test questions have knowledge point labels, the questions under the same concept are grouped together, and the test question scores can be used to evaluate the mastery of existing and potential knowledge points.

[0055] See also Figure 3 , Figure 3This is a flow chart of the third embodiment of the student ability diagnosis method of the present application. Specifically, it may include the following steps:

[0056] Step S31: The computing device inputs the test answer results.

[0057] In the disclosed embodiment, the test answer results can not only be used to form the ability diagnosis results based on the correct / incorrect type labels to obtain the ability diagnosis results based on the students' qualities, but also be used in the disclosed embodiment based on the thought method classification labels to form the ability diagnosis results based on the thought method.

[0058] Step S32: Based on the test answer result, obtain the test question representation, which is in vector form.

[0059] When obtaining the question representation based on the question answering results, the question corresponding to the question answering results can be input into the question representation acquisition model, and the obtained question vector representation is used as the question representation. For a description of the question representation acquisition model, please refer to the above Figure 2 The embodiments are disclosed and will not be described in detail here.

[0060] Step S33: Input the test question representation into the multi-task classification model. The test question representation is processed by a shallow convolution layer, several deep pooling layers and a deep convolution layer to obtain several thought method classification labels.

[0061] The test question representation is input into the multi-task classification model, and the multi-task classification model can output several classification labels of thought methods. The classification labels of thought methods can be determined quickly without relying on manpower.

[0062] Since the shallow network layer can only find common features that are easy to extract, and the thought method classification label is a relatively abstract label, the deep network layer is needed to abstract the thought methods covered by the test questions. In addition, due to the approximate multiplication of the affine matrix of the deep network in the processing of the multi-task classification model, the features represented by the test questions are lost, which makes the multi-task classification model very prone to gradient explosion or diffusion problems. Therefore, the shallow convolution layer of the multi-task classification model is quickly connected to the input of the deep convolution layer to transmit the features represented by the test questions to the deep convolution layer, thereby weakening the impact of the gradient change of the convolution layer and alleviating the diffusion problem.

[0063] The thought method classification label can be various self-defined labels. For example, the thought method classification label can include but is not limited to at least one of number-shape combination, transformation and classification, classification discussion, sampling thought, function and equation, special and general, and probability thought.

[0064] Step S34: Classify labels based on thinking methods to form ability diagnosis results based on thinking methods.

[0065] Similar to the formation of ability diagnosis results based on correct / wrong type labels mentioned above, when forming ability diagnosis results based on idea method classification labels, we can first determine the ratio of test scores of all corresponding idea method classification labels to test standard scores in several test questions; then obtain the average value of the ratios of corresponding idea method classification labels in several test questions, and use the average value as the score of the idea method classification label to form the ability diagnosis result corresponding to each idea method classification label; or we can also count the first sum of test scores of all test questions under the corresponding idea method classification label and the second sum of test standard scores, and use the ratio of the first sum to the second sum as the ability diagnosis result of the corresponding idea method classification label.

[0066] In the above scheme, after the computing device inputs the test answer results, it obtains the test question representation in vector form based on the test question answer results, and then inputs the test question representation into the multi-task classification model. The test question representation is processed by a shallow convolution layer, several deep pooling layers and a deep convolution layer to obtain several idea method classification labels, and then the idea method classification labels can be used to form an ability diagnosis result based on the idea method. Since the shallow convolution layer of the multi-task classification model is quickly connected to the input of the deep convolution layer to transmit the features of the test question representation to the deep convolution layer, the influence of the gradient change of the convolution layer can be weakened, the diffusion problem can be alleviated, and the more abstract idea method classification labels can be classified.

[0067] See also Figure 4 , Figure 4 This is a flowchart of the fourth embodiment of the student ability diagnosis method of the present application. Specifically, it may include the following steps:

[0068] Step S41: The computing device inputs the test answer results.

[0069] Step S42: Based on the test answer result, obtain the test question representation, which is in vector form.

[0070] Step S43: Input the test question representation into the multi-task classification model. The test question representation is processed by a shallow convolutional layer to obtain several classification labels of problem-solving methods.

[0071] Step S44: Based on the problem-solving method classification labels, an ability diagnosis result based on the problem-solving method is formed.

[0072] The problem-solving method classification label can be a variety of custom labels. For example, the problem-solving method classification label includes at least one of the following: variable substitution method, contradiction method, elimination method, number axis intuition method, term splitting and elimination method, and threading method. Since the problem-solving method classification label is a label with a relatively clear boundary, it can be obtained by using the shallow convolution layer of the multi-task classification model. Figure 3Similar to the ability diagnosis results based on thought methods, when forming the ability diagnosis results based on problem-solving methods, the question representation is also input into the multi-task classification model to obtain several problem-solving method classification labels. However, unlike the ability diagnosis results based on thought methods, the question representation can obtain the problem-solving method classification labels after being processed by the shallow convolution layer. That is, when the question representation is input into the multi-task classification model, the question representation is processed by the shallow convolution layer to obtain several problem-solving method classification labels, and after the question representation passes through the shallow convolution layer, it continues to be processed by several deep pooling layers and deep convolution layers to obtain several thought method classification labels.

[0073] Similar to the formation of ability diagnosis results based on correct / incorrect type labels mentioned above, when forming ability diagnosis results based on problem-solving methods based on problem-solving method classification labels, you can first determine the ratio of all test scores of corresponding problem-solving method classification labels to the test standard scores in several test questions; then obtain the average value of the ratios of corresponding problem-solving method classification labels in several test questions, and use the average value as the score of the problem-solving method classification label to form the ability diagnosis result corresponding to each problem-solving method classification label; or you can also count the first sum of the test scores of all test questions under the corresponding problem-solving method classification label and the second sum of the test standard scores, and use the ratio of the first sum to the second sum as the ability diagnosis result of the corresponding problem-solving method classification label.

[0074] Through the above scheme, since the classification labels of problem-solving methods are labels with relatively clear boundaries and the shallow network layer can discover common and easy-to-extract features, the classification labels of problem-solving methods can be quickly obtained through processing by the shallow convolutional layer of the multi-task classification model.

[0075] See also Figure 5 , Figure 5 This is a flowchart of the fifth embodiment of the student ability diagnosis method of the present application. Specifically, it may include the following steps:

[0076] Step S51: The computing device inputs the test question answering results and the corresponding answering time, and the test question answering results and the comparative answering time.

[0077] When the computing device inputs the test answer result, it also inputs the answer time corresponding to the test answer result, several comparative test answer results and their comparative answer times, so as to perform comparative analysis using the comparative test answer results. The comparative test answer results and their comparative answer times can be obtained in any way and from any source. For example, the comparative test answer results can be test questions belonging to the same exam as the test answer result, or test questions of students in the same class as the student corresponding to the test answer result, etc.

[0078] Step S52: Use the test question similarity matching algorithm to obtain the similarity value between the test question answer result and the answer results of several comparison test questions, and make a first mark on the test question answer results whose similarity value exceeds the preset similarity threshold, or use the average value of the comparison answer time / or a preset ratio of the average value as the answer time comparison value, and make a second mark on the test question answer results whose answer time exceeds the answer time comparison value.

[0079] The learning attitude mainly counts the degree of seriousness of students in forming the test answer results and whether they cheat, so that teachers can understand the students' situation in time and make timely corrections. The test question similarity matching algorithm is used to obtain the similarity value between the test question answer result and the answer results of several comparative test questions, and the test question answer results whose similarity value exceeds the preset similarity threshold are first marked to determine whether the student has plagiarized, or the average value of the comparative answer time / or the preset ratio of the average value is used as the answer time comparison value, and the test question answer results whose answer time exceeds the answer time comparison value are second marked to determine whether the student has not answered seriously. The test question similarity matching algorithm can be any algorithm implementation method in the optical character recognition application. Before using the test question similarity matching algorithm to obtain the similarity value between the test question answer result and the answer results of several comparative test questions, the test question answer result can be segmented to obtain the segmentation result. If the number of segmentation results is less than the preset number, the test question answer result can be discarded. In order to improve the accuracy of ability diagnosis, the test answer results can be marked only when the number of test answer results whose similarity values ​​exceed the preset similarity threshold and the number of test answer results whose answer time exceeds the answer time comparison value exceeds the preset number of test questions. The above-mentioned preset similarity threshold, preset ratio, preset number, and preset number of test questions can all be customized. In actual scenarios, the first mark and / or the second mark can be performed as needed. For example, the test answer result may be the test answer result of the answer question, or it may be the test answer result of the multiple-choice question or the fill-in-the-blank question. Since the test answer results of the multiple-choice question and the fill-in-the-blank question are more likely to be similar, the probability of the test answer result being marked increases. Therefore, for the test answer results of the multiple-choice question and the fill-in-the-blank question, only the second mark can be performed, while the test answer result of the answer question can be the first mark and the second mark.

[0080] Step S53: Based on the marked test answer results, obtain the ability diagnosis results based on the learning attitude.

[0081] Based on the marked test answer results, the ability diagnosis results based on learning attitude are obtained, for example, based on the marking type and / or number of markings of the marked test answer results, points are deducted from the current learning attitude score. In order to more flexibly determine the ability diagnosis results based on learning attitude, the marking forgetting can be performed at a certain period, and the ability diagnosis results based on learning attitude can be updated in real time.

[0082] In a public embodiment, when a computing device inputs the test answer results of several test questions, it can perform a student ability diagnosis based on multiple test questions, so that when an ability diagnosis result based on learning attitude is obtained based on the marked test answer results, a first number of test answer results with a first mark can be counted, and if it is determined that the first number exceeds a first preset number, the difference between the current learning attitude score and the first preset score is used as the ability diagnosis result based on learning attitude; or, a second number of test answer results with a second mark can be counted, and if it is determined that the second number exceeds a second preset number, the difference between the current learning attitude score and the second preset score is used as the ability diagnosis result based on learning attitude, thereby accumulating the number of marks, and when the number of marks is large, points are deducted from the current learning attitude score, and the ability diagnosis result based on learning attitude is continuously updated.

[0083] Through the above scheme, after the computing device inputs the test answer results and the corresponding answer time, and the answer results of several comparative test questions and their comparative answer times, the test answer results that are similar to the comparative test questions and the test answer results with shorter answer times are marked respectively, so as to evaluate the seriousness of the students in forming the test answer results and whether they cheat, so that the teacher can understand the students' situation in time and make timely corrections.

[0084] In the above public embodiments, the ability diagnosis results based on student literacy, the ability diagnosis results based on ideas and methods, the ability diagnosis results based on problem-solving methods, the ability diagnosis results based on weak knowledge points, and the ability diagnosis results based on learning attitudes can be arbitrarily combined to construct a multi-task model for multi-dimensional comprehensive ability assessment, output students' comprehensive abilities in five aspects, including weak knowledge points, ideas and methods, student literacy, problem-solving methods, and learning attitudes, to achieve multi-dimensional comprehensive ability diagnosis results. The above ability diagnosis results are formed based on various labels. For example, the ability diagnosis results based on student literacy are formed based on correct / wrong type labels, the ability diagnosis results based on problem-solving methods are formed based on problem-solving method classification labels, the ability diagnosis results based on ideas and methods are formed based on idea classification labels, and the ability diagnosis results based on weak knowledge points are formed based on knowledge point labels. The public embodiments of the above student ability diagnosis method can perform ability diagnosis on one or several target students, which is particularly suitable for application scenarios such as smart classrooms. Students can quickly enter the test answer results through online answering, and feedback various ability diagnosis results in real time. In the above-mentioned public embodiments, data such as test answer results, test questions, test scores, test standard scores, reference questions, test questions, related questions, answer time, etc. can also be obtained from the test database, so as to facilitate the diagnosis of students' abilities based on their historical answer situations.

[0085] In an actual scenario, for the ability diagnosis results of one dimension, the ability diagnosis results of the corresponding labels can be sorted in descending order, and the ability diagnosis results corresponding to the labels of the first ranking number are used as the ability diagnosis results of the corresponding dimension, wherein the first ranking number can be customized, such as 1, 2, 3, etc. For the ability diagnosis results of multiple dimensions, the ability diagnosis results corresponding to the labels of multiple dimensions can also be sorted in descending order, and the ability diagnosis results of the second ranking number are used as the final ability diagnosis results, wherein the second ranking number can be customized, such as 1, 2, 3, etc. When the computing device inputs the test answer results of several test questions of the target student and several comparison students, the target student and the comparison student can both perform multi-dimensional ability diagnosis, thereby obtaining the average value of the ability diagnosis results of several comparison students corresponding to each label, so as to take the difference between the ability diagnosis result of the target student and the average value of the ability diagnosis result as the ability diagnosis result corresponding to the label of the target student. In the above manner, the ability diagnosis results of the comparison students are summed and averaged, which can reduce the random error caused by data sparseness, and then the ability diagnosis results of the target student and the comparison student are correspondingly subtracted, and the target student and the comparison student are compared and sorted, so as to avoid the direct diagnosis of the target student, which may have large accidental errors. In addition, the ability diagnosis results based on learning attitude are obtained based on the answer results of the marked test questions, which can reflect situations such as students plagiarizing or answering questions carelessly. Therefore, if the ability diagnosis results of the above labels are relatively good, but the ability diagnosis results based on learning attitude are relatively poor, there may be a possibility of plagiarism; or if the ability diagnosis results of the above labels are relatively poor, and the ability diagnosis results based on learning attitude are relatively poor, there may be a possibility of answering questions carelessly. To reflect these situations, the ability diagnosis results based on learning attitude can be separately reflected as a student ability diagnosis situation, and it will not participate in the above descending order.

[0086] See also Figure 6 , Figure 6 Schematic diagram of a framework of an embodiment of a computing device 60 of the present application. The computing device 60 includes an input module 61, a disassembly module 62, a determination module 63, and a result formation module 64. The input module 61 is used to input the test answer result; the disassembly module 62 is used to disassemble the test answer result into steps to obtain a number of answer steps; the determination module 63 is used to analyze the answer steps using a network model and determine the correct / wrong type label; the result formation module 64 is used to form the ability diagnosis result based on the correct / wrong type label.

[0087] In the above scheme, after the computing device 60 inputs the test answer result into the input module 61, the disassembly module 62 disassembles the test answer result into steps to obtain a number of answer steps, and then the determination module 63 uses the network model to analyze the answer steps and determine the correct / incorrect type label. Finally, the result formation module 64 forms the ability diagnosis result based on the correct / incorrect type label. The whole process is completed by the computing device 60, so that the student's ability can be intelligently diagnosed.

[0088] In some disclosed embodiments, the computing device 60 determines a module 63 for analyzing answering steps using a network model to determine correct / incorrect type labels, and is also used to determine correct / incorrect type labels for scoring points based on the answering steps, and / or to determine correct / incorrect type labels for step relationships based on the relationships between different answering steps.

[0089] Therefore, analysis can be performed on each answering step, or on the relationship between different answering steps, to increase the depth of analysis of the test answering results.

[0090] In some disclosed embodiments, when the determination module 63 is used to determine the correct / incorrect type label of the scoring point based on the answering steps, it is also used to extract the scoring points of the answering steps using a scoring point extraction model to obtain at least one detection scoring point; it is judged whether the detection scoring point matches the standard scoring point, wherein there is a scoring point mapping relationship between the standard scoring point and the correct / incorrect type label; if so, the correct / incorrect type label of the detection scoring point is determined based on the scoring point mapping relationship; and / or, when the determination module 63 is used to determine the correct / incorrect type label of the step relationship based on the relationship between different answering steps, it is also used to parse the detection step relationship between different answering steps using a paragraph structure parsing model; it is judged whether the detection step relationship matches the standard step relationship, wherein there is a step relationship mapping relationship between the standard step relationship and the correct / incorrect type label; if so, the correct / incorrect type label of the detection step relationship is determined based on the step relationship mapping relationship.

[0091] Therefore, the scoring point extraction model is used to determine whether the test scoring points match the standard scoring points. The test scoring points are compared with the correct standard scoring points by word segmentation, and the correct / incorrect type labels corresponding to the students can be quickly output. Similarly, the detection step relationship between different answer steps is analyzed by the paragraph structure parsing model, and the correct / incorrect type labels of the detection step relationship can be quickly determined, which can not only save manpower and avoid omissions in human judgment, but also improve diagnostic efficiency.

[0092] In some disclosed embodiments, error type labels for detection scoring points include: weak computing power, lack of data analysis dimension; and / or, detection step relationships include conditions and derivations, and error type labels for detection step relationships include condition usage errors and derivation errors. The determination module 63 is used to utilize the paragraph structure parsing model to parse the detection step relationships between different answering steps, and is also used to utilize the paragraph structure parsing model's beam search to determine the detection step relationship with the highest conditional probability.

[0093] Therefore, when the paragraph structure parsing model parses the detection step relationship between different answering steps, it combines beam search to find the optimal path, finds the sequence with the largest reasonable conditional probability of the corresponding answering step, and infers the next most reasonable answering step, thereby completing the parsing of the detection step relationship.

[0094] In some disclosed embodiments, when the input module 61 of the computing device 60 is used to input the test answer results, the computing device 60 is also used to input the test answer results of several test questions; when the result forming module 64 of the computing device 60 is used to form an ability diagnosis result based on the correct / incorrect type label, it is also used to determine the ratio of all test scores corresponding to the incorrect type labels to the test standard score in several test questions; obtain the average value of the ratios corresponding to the incorrect type labels in several test questions, use the average value as the score of the incorrect type label, and form an ability diagnosis result corresponding to each incorrect type label.

[0095] Therefore, using the correct / incorrect type labels, test scores and test standard scores, under each error type label, the average of the ratios of all test questions belonging to the error type label is calculated to obtain the score corresponding to each error type label, forming an ability diagnosis result, and thus determining the student's ability by combining the test answer results of multiple questions.

[0096] In some disclosed embodiments, when the input module 61 of the computing device 60 is used to input the test answer results, the computing device 60 is also used to input the test answer results of several test questions of the target student and several comparison students; when the result forming module 64 of the computing device 60 is used to form the ability diagnosis result based on the correct / incorrect type label, it is also used to obtain the average value of the ability diagnosis results of several comparison students corresponding to each error type label; the difference between the ability diagnosis result of the target student and the average value of the ability diagnosis result is used as the ability diagnosis result corresponding to the error type label of the target student.

[0097] Therefore, when diagnosing the target students' abilities, the ability diagnosis results of the comparison students are included, and the difference between the target students' ability diagnosis results and the average value of the comparison students' ability diagnosis results is used as the final ability diagnosis result of the target students, which can reduce the random errors caused by data sparsity.

[0098] In some public embodiments, the disassembly module 62 is used to decompose the test answer results into steps. When a number of answer steps are obtained, the disassembly module 62 is also used to split the test answer results into a number of answer steps based on preset symbols in the test answer results; and segment the answer steps using a preset word segmentation set to obtain a number of answer steps after segmentation.

[0099] Therefore, the test answer results can be broken down into steps through preset symbols, and the answer steps can be segmented using preset word segmentation sets, which can make the word segmentation processing more flexible and closer to the actual scenario.

[0100] In some disclosed embodiments, after the input module 61 of the computing device 60 is used to input the test answer results, the computing device 60 is also used to obtain a test question representation based on the test question answer results, and the test question representation is in vector form; the test question representation is input into a multi-task classification model, and the test question representation is processed by a shallow convolution layer, several deep pooling layers and a deep convolution layer to obtain several idea method classification labels, wherein the shallow convolution layer of the multi-task classification model is quickly connected to the input of the deep convolution layer to transfer the features of the test question representation to the deep convolution layer; based on the idea method classification labels, an ability diagnosis result based on the idea method is formed; wherein the idea method classification labels include at least one of the combination of numbers and shapes, transformation and classification, classification discussion, sampling ideas, functions and equations, special and general, and probability ideas.

[0101] Therefore, the shallow convolutional layer of the multi-task classification model is quickly connected to the input of the deep convolutional layer to transfer the features represented by the test questions to the deep convolutional layer, thereby weakening the influence of the gradient change of the convolutional layer, alleviating the diffusion problem, and realizing the classification of more abstract thought method classification labels.

[0102] In some disclosed embodiments, after the input module 61 of the computing device 60 is used to input the test answer results, the computing device 60 is also used to obtain a test question representation based on the test question answer results, and the test question representation is in vector form; the test question representation is input into a multi-task classification model, and the test question representation is processed by a shallow convolutional layer to obtain a number of problem-solving method classification labels; based on the problem-solving method classification labels, an ability diagnosis result based on the problem-solving method is formed; wherein the problem-solving method classification labels include at least one of the substitution method, the proof by contradiction, the elimination method, the number axis intuition method, the term splitting and elimination method, and the threading method.

[0103] Therefore, the classification labels of problem-solving methods are labels with relatively clear boundaries, and the shallow network layer can discover common features that are easy to extract. Therefore, the test question representation can be processed by the shallow convolutional layer of the multi-task classification model to obtain the classification labels of problem-solving methods.

[0104] In some disclosed embodiments, the computing device 60 is also used to input a number of reference test questions and their corresponding reference test question scores, and a number of test test questions including knowledge point labels into the computing device 60; obtain a reference test question representation of each reference test question based on the number of reference test questions and their corresponding reference test question scores, and obtain a test test question representation of each test question based on the number of test questions; input all reference test question representations and test test question representations into a score prediction model to obtain a test test question score for each test question; and form an ability diagnosis result based on weak knowledge points based on the test question scores and the corresponding knowledge point labels of the test questions.

[0105] Therefore, unlike mapping wrong questions with knowledge points, counting the number of wrong questions or scores, and directly obtaining weak knowledge points, the statistical method is too rigid, and the number of questions affects the coverage, resulting in too much accidental factors in the diagnosis results. In this case, all reference question representations and test question representations are input into the score prediction model to obtain the test question score of each test question, and then based on the test question score and the corresponding knowledge point label of the test question, an ability diagnosis result based on weak knowledge points is formed. On the one hand, students do not need to actually do a large number of reference questions, which can reduce the number of reference questions. On the other hand, because the test questions carry knowledge point labels, questions under the same concept are grouped together, and the test question scores can be used to evaluate the mastery of existing and potential knowledge points.

[0106] In some disclosed embodiments, when the computing device 60 is used to obtain a reference question representation of each reference question based on a number of reference questions and their corresponding reference question scores, the computing device 60 is also used to input the reference questions into a question representation acquisition model to obtain a reference question vector representation of the reference question; normalize the reference question scores to obtain a reference score representation of the reference question scores; and use the product of the corresponding reference question vector representation and the reference score representation as the reference question representation; the computing device 60 is used to obtain a test question representation of each test question based on a number of test questions, and the computing device 60 is also used to input the test questions into a question representation acquisition model to obtain a test question representation of the test question.

[0107] Therefore, since the reference score representation obtained by incorporating the reference test question score when obtaining the reference test question representation, the reference score representation reflects to a certain extent the student's mastery of the reference test question. Therefore, the reference test question representation is the preliminary result of the personal ability assessment and forms the initial representation of the student's personal learning.

[0108] In some disclosed embodiments, the computing device 60 is used to normalize the reference test question scores respectively to obtain the reference score representation of the reference test question scores, and is also used to obtain several related test question scores of several related test questions associated with the reference test question; to obtain the mean of the several related test question scores, to determine the difference between the reference test question score and the mean, and to obtain the standard deviation of the reference test question score and the several related test question scores, and to express the reference score as the quotient of the difference and the standard deviation; or, to obtain a first difference between the reference test question score and the smallest related test question score, and to obtain a second difference between the largest related test question score and the reference test question score, and to express the reference score as the quotient of the first difference and the second difference.

[0109] Therefore, compared with directly using the reference test question scores to obtain the reference test question representation, the reference test question scores have a greater impact on the vector. By normalizing the reference test question scores, not only can the associated test question scores be used to reduce data randomness, but the normalized processing results are also more conducive to the representation of the vector.

[0110] In some disclosed embodiments, the computing device 60 is used to input all reference question representations and test question representations into a score prediction model to obtain a test question score for each test question. It is also used to match the reference question representations and the test question representations, respectively determine the reference question with the greatest correlation with each test question, and use the reference question score of the reference question with the greatest correlation to predict the test question score of the corresponding test question.

[0111] Therefore, since the test questions carry knowledge point labels, questions under the same concept are grouped together, and the reference question with the greatest correlation to each test question is determined. The reference question score of the reference question with the greatest correlation is used to predict the test question score of the corresponding test question. The test question score can be used to evaluate the mastery of existing and potential knowledge points.

[0112] In some disclosed embodiments, the test questions include standard scores for the test questions. The computing device 60 is used to form an ability diagnosis result based on weak knowledge points based on the test question scores and the corresponding knowledge point labels of the test questions. The computing device 60 is also used to determine the ratio of the test question scores to the test question standard scores of all test questions; obtain the average value of the ratios of the test questions with corresponding knowledge point labels, use the average value as the score of the knowledge point label, and form an ability diagnosis result corresponding to each knowledge point label.

[0113] Therefore, by predicting students' weak knowledge points through test question scores, the number of reference questions can be reduced without students having to actually do a large number of reference questions.

[0114] In some disclosed embodiments, when the input module 61 of the computing device 60 is used to input the test answer result, the answer time corresponding to the test answer result, several comparative test answer results and their comparative answer times are also input; the computing device 60 is used to use the test question similarity matching algorithm to obtain the similarity value between the test question answer result and several comparative test question answer results, and to make a first mark on the test question answer result whose similarity value exceeds a preset similarity threshold, or to use the average value of the comparative answer time / or a preset ratio of the average value as the answer time comparison value, and to make a second mark on the test question answer result whose answer time exceeds the answer time comparison value; based on the marked test question answer result, an ability diagnosis result based on learning attitude is obtained.

[0115] Therefore, after the computing device 60 inputs the test answer results and the corresponding answer time, several comparison test answer results and their comparison answer times, the test answer results that are more similar to the comparison test questions and the test answer results with shorter answer times are marked respectively, so as to evaluate the seriousness of the test answer results and whether there is cheating, so that the teacher can understand the student situation in time and make timely corrections.

[0116] In some disclosed embodiments, when the computing device 60 inputs the test answer results of several test questions, when obtaining the ability diagnosis result based on the learning attitude based on the marked test answer results, it is also used to count the first number of test answer results with the first mark, and if it is determined that the first number exceeds the first preset number, the difference between the current learning attitude score and the first preset score is used as the ability diagnosis result based on the learning attitude; or, count the second number of test answer results with the second mark, and if it is determined that the second number exceeds the second preset number, the difference between the current learning attitude score and the second preset score is used as the ability diagnosis result based on the learning attitude.

[0117] Therefore, the number of marks is accumulated, and when the number of marks is large, points are deducted from the current learning attitude score, and the ability diagnosis results based on learning attitude are continuously updated.

[0118] In some disclosed embodiments, the result forming module 64 of the computing device 60 is used to form an ability diagnosis result based on the correct / incorrect type label, and is also used to obtain and utilize at least one of the ability diagnosis result based on the idea method, the ability diagnosis result based on the problem-solving method, the ability diagnosis result based on the weak knowledge point, and the ability diagnosis result based on the learning attitude to form a comprehensive ability diagnosis result.

[0119] Therefore, the dimensions of student ability diagnosis can be flexibly selected to achieve multi-dimensional ability diagnosis.

[0120] See also Figure 7 , Figure 7: is a schematic diagram of a framework of an embodiment of an electronic device 70 of the present application. The electronic device 70 includes a memory 71 and a processor 72 coupled to each other, the memory 71 stores program instructions, and the processor 72 is used to execute the program instructions to implement the steps in any of the above-mentioned student ability diagnosis method embodiments. Specifically, the electronic device 70 may include but is not limited to: a desktop computer, a laptop computer, a server, a mobile phone, a tablet computer, etc., which are not limited here.

[0121] Specifically, the processor 72 is used to control itself and the memory 71 to implement the steps in any of the above-mentioned student ability diagnosis method embodiments. The processor 72 can also be called a CPU (Central Processing Unit). The processor 72 may be an integrated circuit chip with signal processing capabilities. The processor 72 can also be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. In addition, the processor 72 can be implemented by an integrated circuit chip.

[0122] In the above scheme, after the computing device inputs the test answer results, the test answer results are broken down into steps to obtain several answer steps, and then the answer steps are analyzed using a network model to determine the correct / incorrect type labels. Finally, based on the correct / incorrect type labels, an ability diagnosis result is formed. The entire process is completed by the computing device, thereby being able to intelligently diagnose students' abilities.

[0123] See also Figure 8 , Figure 8 The schematic diagram is a schematic diagram of a computer-readable storage medium 80 of the present application. The computer-readable storage medium 80 stores program instructions 81 that can be executed by a processor, and the program instructions 81 are used to implement the steps in any of the above-mentioned student ability diagnosis method embodiments.

[0124] In the above scheme, after the computing device inputs the test answer results, the test answer results are broken down into steps to obtain several answer steps, and then the answer steps are analyzed using a network model to determine the correct / incorrect type labels. Finally, based on the correct / incorrect type labels, an ability diagnosis result is formed. The entire process is completed by the computing device, thereby being able to intelligently diagnose students' abilities.

[0125] In some embodiments, the functions or modules included in the device provided by the embodiments of the present disclosure can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0126] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0127] In the several embodiments provided in the present application, it should be understood that the disclosed methods and devices can be implemented in other ways. For example, the device implementation described above is only schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0128] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0129] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0130] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of each implementation method of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

Claims

1. A method for diagnosing student ability, characterized in that: The method comprises: The computing device inputs the test answer result; The computing device decomposes the test answer result into steps to obtain a plurality of answer steps; The computing device uses a network model to analyze the answering steps and determine a correct / incorrect type label; The computing device forms a capability diagnosis result based on the correct / incorrect type label; The computing device uses a network model to analyze the answering steps and determine a correct / wrong type label, including: Utilizing the paragraph structure analysis model to analyze the detection step relationship between the different answering steps; Determine whether the detection step relationship matches the standard step relationship, wherein there is a step relationship mapping relationship between the standard step relationship and the correct / incorrect type label; If so, the correct / incorrect type label of the detection step relationship is determined based on the step relationship mapping relationship.

2. The method according to claim 1, characterized in that The computing device uses a network model to analyze the answering steps and determine a correct / wrong type label, and further includes: Based on the answering steps, a correct / incorrect type label of the scoring point is determined.

3. The method according to claim 2, characterized in that The step of determining the correct / incorrect type label of the scoring point based on the answering step includes: Extracting the score points of the answering step using a score point extraction model to obtain at least one detection score point; Determining whether the detection score point matches the standard score point, wherein there is a score point mapping relationship between the standard score point and the correct / incorrect type label; If so, the correct / incorrect type label of the detection score point is determined based on the score point mapping relationship.

4. The method according to claim 3, characterized in that The error type labels of the detection score points include: weak computing power and lack of data analysis dimensions; And / or, the detection step relationship includes conditions and deductions, the error type labels of the detection step relationship include condition usage errors and deduction errors, and the use of the paragraph structure parsing model to parse the detection step relationship between different answering steps includes: using the beam search of the paragraph structure parsing model to determine the detection step relationship with the highest conditional probability.

5. The method according to claim 1, characterized in that The computing device inputting the test answer results includes: the computing device inputting the test answer results of a plurality of test questions; The computing device forms a capability diagnosis result based on the correct / incorrect type label, including: Determine the ratio of all test scores corresponding to the error type labels to the test standard score in the plurality of test questions; The average value of the ratios corresponding to the error type labels in the several test questions is obtained, and the average value is used as the score of the error type label to form an ability diagnosis result corresponding to each error type label.

6. The method according to claim 5, characterized in that The computing device inputting the test answer results includes: the computing device inputting the test answer results of a plurality of test questions of a target student and a plurality of comparison students; The computing device forms a capability diagnosis result based on the correct / incorrect type label, including: Obtaining an average value of the ability diagnosis results corresponding to each error type label of the plurality of comparison students; The difference between the ability diagnosis result of the target student and the average value of the ability diagnosis results is used as the ability diagnosis result corresponding to the error type label of the target student.

7. The method according to claim 1, characterized in that The computing device decomposes the test answer result into steps to obtain a plurality of answer steps, including: Splitting the test answer result into a plurality of answer steps using preset symbols in the test answer result as demarcations; The answering steps are segmented using a preset segmentation set to obtain a plurality of the answering steps after segmentation.

8. The method according to claim 1, characterized in that After the computing device inputs the test answer result, the method further includes: Based on the test answer result, obtain a test question representation, wherein the test question representation is in vector form; Input the question representation into a multi-task classification model, the question representation is processed by a shallow convolution layer, several deep pooling layers and a deep convolution layer to obtain several thought method classification labels, wherein the shallow convolution layer of the multi-task classification model is quickly connected to the input of the deep convolution layer to transfer the features of the question representation to the deep convolution layer; Based on the classification labels of the thought methods, a capability diagnosis result based on the thought methods is formed; Among them, the classification labels of the thought methods include at least one of the combination of numbers and shapes, transformation and classification, classification discussion, sampling ideas, functions and equations, special and general, and probability ideas.

9. The method according to claim 1, characterized in that: After the computing device inputs the test answer result, the method further includes: Based on the test answer result, obtain a test question representation, wherein the test question representation is in vector form; Inputting the test question representation into a multi-task classification model, the test question representation is processed by a shallow convolutional layer to obtain several classification labels of problem-solving methods; Based on the problem-solving method classification labels, forming an ability diagnosis result based on the problem-solving method; Among them, the classification labels of problem-solving methods include at least one of the substitution method, proof by contradiction, elimination method, number axis intuitive method, term splitting and elimination method, and threading the needle method.

10. The method according to claim 1, characterized in that The method further comprises: The computing device inputs a number of reference test questions and their corresponding reference test question scores, and a number of test test questions including knowledge point labels; Based on the reference test questions and their corresponding reference test question scores, respectively obtaining a reference test question representation of each of the reference test questions, and based on the test test questions, respectively obtaining a test test question representation of each of the test test questions; Inputting all the reference question representations and the test question representations into a score prediction model to obtain a test question score for each of the test questions; Based on the test question scores and the corresponding knowledge point labels of the test questions, an ability diagnosis result based on weak knowledge points is formed.

11. The method according to claim 10, characterized in that The obtaining, based on the reference test questions and their corresponding reference test question scores, a reference test question representation of each reference test question, comprises: Input the reference test questions into a test question representation acquisition model respectively to obtain a reference test question vector representation of the reference test question; Normalizing the reference test question scores respectively to obtain reference score representations of the reference test question scores; The product of the corresponding reference test question vector representation and the reference score representation is used as the reference test question representation; The step of obtaining a test question representation of each of the test questions based on the plurality of test questions comprises: The test questions are respectively input into the question representation acquisition model to obtain the test question representation of the test question.

12. The method according to claim 11, characterized in that The normalizing the reference test question scores respectively to obtain reference score representations of the reference test question scores includes: Obtaining scores of several related test questions associated with the reference test question; Obtain the mean of the scores of the several related test questions, determine the difference between the reference test question score and the mean, and obtain the standard deviation of the reference test question score and the scores of the several related test questions, and express the quotient of the difference and the standard deviation as the reference score; or obtain a first difference between the reference test question score and the smallest related test question score, and obtain a second difference between the largest related test question score and the reference test question score, and express the quotient of the first difference and the second difference as the reference score.

13. The method according to claim 10, characterized in that The step of inputting all the reference question representations and the test question representations into a score prediction model to obtain a test question score for each test question includes: The reference question representation and the test question representation are matched, the reference question with the greatest correlation with each of the test questions is respectively determined, and the test question score corresponding to the test question is predicted using the reference question score of the reference question with the greatest correlation.

14. The method according to claim 10, characterized in that The test question includes a standard score of the test question, and the ability diagnosis result based on weak knowledge points is formed based on the test question score and the corresponding knowledge point label of the test question, including: Determining a ratio of the test question score to the test question standard score for all the test questions; The average value of the ratios of the test questions corresponding to the knowledge point labels is obtained, and the average value is used as the score of the knowledge point label to form an ability diagnosis result corresponding to each knowledge point label.

15. The method according to claim 1, characterized in that When the computing device inputs the test question answering result, the computing device also inputs the answering time corresponding to the test question answering result, a number of comparative test question answering results and their comparative answering times; the method further includes: Using a test question similarity matching algorithm to obtain a similarity value between the test question answer result and the several comparison test question answer results, a first mark is made for the test question answer result whose similarity value exceeds a preset similarity threshold, or an average value of the comparison answer time / or a preset ratio of the average value is used as an answer time comparison value, and a second mark is made for the test question answer result whose answer time exceeds the answer time comparison value; Based on the marked test answer results, an ability diagnosis result based on learning attitude is obtained.

16. The method according to claim 15, characterized in that When the computing device inputs the test answer results of several test questions, the ability diagnosis results based on the marked test answer results are obtained, including: Counting a first number of the test answer results of the first mark, and determining that the first number exceeds a first preset number, and taking the difference between the current learning attitude score and the first preset score as the ability diagnosis result based on the learning attitude; or A second number of the test answer results of the second mark is counted, and if it is determined that the second number exceeds a second preset number, the difference between the current learning attitude score and the second preset score is used as the ability diagnosis result based on learning attitude.

17. The method according to claim 1, characterized in that The computing device forms a capability diagnosis result based on the correct / incorrect type label, further comprising: Obtain and utilize at least one of the ability diagnosis results based on ideas and methods, the ability diagnosis results based on problem-solving methods, the ability diagnosis results based on weak knowledge points, and the ability diagnosis results based on learning attitudes to form a comprehensive ability diagnosis result.

18. A computing device, characterized in that The computing device comprises: An input module is used to input the test answer results; A disassembly module is used to disassemble the test question answer result into steps to obtain a plurality of answer steps; A determination module, used to analyze the answering steps using a network model to determine a correct / incorrect type label; A result forming module, used for forming a capability diagnosis result based on the correct / incorrect type label; The computing device uses a network model to analyze the answering steps and determine a correct / wrong type label, including: Utilizing the paragraph structure analysis model to analyze the detection step relationship between the different answering steps; Determine whether the detection step relationship matches the standard step relationship, wherein there is a step relationship mapping relationship between the standard step relationship and the correct / incorrect type label; If so, the correct / incorrect type label of the detection step relationship is determined based on the step relationship mapping relationship.

19. An electronic device, characterized in that: It comprises a memory and a processor coupled to each other, wherein the memory stores program instructions, and the processor is used to execute the program instructions to implement the student ability diagnosis method described in any one of claims 1 to 17.

20. A computer-readable storage medium, characterized in that: Program instructions that can be executed by a processor are stored, and the program instructions are used to implement the student ability diagnosis method described in any one of claims 1 to 17.

Images