CN113793112A - A method, system and medium for automatic correction of programming operations - Google Patents

Abstract

The invention discloses a method, system and medium for automatic correction of programming jobs. The method includes: acquiring programming jobs to be corrected; inputting the programming jobs into a preset neural network model, and calculating and outputting characteristic data of the programming jobs ; Carry out feature matching between the feature data of the programming job and the correct answer template set, and output corresponding confidence data; confirm the correction result of the programming job according to the confidence data. In the embodiment of the present invention, the feature data of the programming job is calculated by the neural network model, and the corresponding confidence data is obtained after feature matching between the feature data and the correct answer template. Efficient automatic correction improves the efficiency of homework correction in the process of programming teaching.

Description

A method, system and medium for automatic correction of programming operation

technical field

The present invention relates to the technical field of graphical programming, in particular to a method, system and medium for automatic correction of programming operations.

Background technique

In the current online children's programming courses, students' after-school programming assignments mainly rely on manual corrections, mainly because the graphical programming assignments are composed of graphical building blocks representing various variables and operations. The degree of freedom is high, and a job usually corresponds to multiple implementation methods. The existing automatic correction only provides a few correct answers for accurate comparison, which makes the efficiency of automatic correction low, and it is difficult to meet the needs of efficient automatic correction of various programming jobs.

SUMMARY OF THE INVENTION

In view of the above-mentioned deficiencies of the prior art, the purpose of the present invention is to provide a method, system and medium for automatic correction of programming operations, so as to improve the efficiency of automatic correction of programming operations.

The technical scheme of the present invention is as follows:

An automatic correction method for programming jobs, which comprises the following steps:

Get programming assignments to be corrected;

Inputting the programming job into a preset neural network model, and calculating and outputting characteristic data of the programming job;

Perform feature matching on the feature data of the programming job with the correct answer template set, and output corresponding confidence data;

The correction result of the programming job is confirmed according to the confidence data.

In the described programming job automatic correction method, the feature matching is performed between the feature data of the programming job and the correct answer template set, and the corresponding confidence data is output, including:

Input the feature data of the programming job into a preset matching model, and output the confidence data between the programming job and the preset template category, the matching model is based on all the obtained correct answer template sets. The correct answer template is trained.

In the method for automatic correction of programming jobs, before the feature data of the programming job is input into the preset matching model, the method further includes:

Classify all the correct answer templates that have been obtained in the correct answer template set, and mark each correct answer template as a corresponding preset template category;

Input each correct answer template into the neural network model respectively, and calculate and output the characteristic training data of each correct answer template;

The matching model is constructed, and the matching model is trained according to the preset template category of each correct answer template and the feature training data, until the matching model satisfies preset convergence conditions.

In the automatic correction method for programming jobs, the confirming the correction result of the programming job to be corrected according to the confidence data includes:

If there is confidence data greater than or equal to the confidence threshold, the correction result of the programming job is qualified;

If the confidence data are all less than the confidence threshold, the correction result of the programming job is unqualified;

classifying the programming jobs whose correction result is unqualified into the set of jobs to be processed;

Cluster processing is performed on the set of jobs to be processed, new correct answer templates are extracted and classified into the set of correct answer templates, and feature matching is continued until all correct answers are matched from all programming jobs to be corrected.

In the method for automatic correction of programming jobs, before the inputting the programming job into a preset neural network model, and before calculating and outputting the characteristic data of the programming job, the method further includes:

The programming job is serialized according to preset rules.

In the method for automatic correction of programming jobs, the serializing processing of the programming jobs according to preset rules includes:

Get the building block code block combination in the programming assignment;

Analyze the building block code block combination in a preset order to obtain the building block type and building block data corresponding to each building block code block;

A corresponding serialized array is generated after serializing according to the building block type and building block data according to preset rules.

In the method for automatic correction of programming jobs, the combination of building block code blocks is analyzed in a preset order to obtain the building block type and building block data corresponding to each building block code block, including:

Obtain the splicing sequence of the building block code block combination;

Each building block code block is sequentially parsed according to the splicing sequence to obtain building block types and building block data arranged according to the splicing sequence.

In the method for automatic correction of programming jobs, after serializing according to the building block type and building block data according to preset rules, a corresponding serialized array is generated, including:

The building block type and building block data are respectively converted into corresponding encoded data according to a preset encoding mapping library, and the encoding mapping library is used to store the mapping relationship between the building block type and the building block data and the encoded data;

After encapsulating the encoded data according to the preset data structure, a one-dimensional array corresponding to each building block code block is obtained;

The one-dimensional arrays are superimposed according to the splicing sequence to generate a two-dimensional serialized array corresponding to the combination of the building block code blocks.

Yet another embodiment of the present invention also provides an automatic correction system for programming jobs, the system includes at least one processor; and,

a memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described method for automatic correction of programming jobs.

Another embodiment of the present invention also provides a non-volatile computer-readable storage medium storing computer-executable instructions, the computer-executable instructions being stored by one or more When the processor executes, the one or more processors can be made to execute the above-mentioned method for automatic correction of programming jobs.

Beneficial effects: The present invention discloses a method, system and medium for automatic correction of programming jobs. Compared with the prior art, the embodiment of the present invention calculates characteristic data of programming jobs through a neural network model, and compares the characteristic data with the correct answer template. Corresponding confidence data is obtained after feature matching, and the confidence data obtained from deep learning training and feature matching can be used for efficient automatic correction of programming assignments, which improves the efficiency of assignment correction during programming teaching.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings and embodiments, in which:

1 is a flowchart of a preferred embodiment of a method for automatically correcting programming operations provided by the present invention;

2 is a schematic diagram of a combination of building blocks code blocks in a programming job to be processed in an application embodiment of the method for automatically correcting programming jobs provided by the present invention;

3 is a flowchart of feature data calculation in an application embodiment of the method for automatically correcting programming jobs provided by the present invention;

4 is a flowchart of template matching in an application embodiment of a method for automatically correcting programming jobs provided by the present invention;

FIG. 5 is a schematic diagram of functional modules of a preferred embodiment of the device for automatic correction of programming operations provided by the present invention;

FIG. 6 is a schematic diagram of the hardware structure of a preferred embodiment of the system for automatic correction of programming operations provided by the present invention.

Detailed ways

In order to make the objectives, technical solutions and effects of the present invention clearer and clearer, the present invention will be described in further detail below. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. The embodiments of the present invention are described below with reference to the accompanying drawings.

Please refer to FIG. 1 . FIG. 1 is a flowchart of a preferred embodiment of a method for automatic correction of programming jobs provided by the present invention. As shown in Figure 1, it includes the following steps:

S100. Obtain the programming job to be corrected;

S200, inputting the programming job into a preset neural network model, and calculating and outputting characteristic data of the programming job;

S300, feature matching the feature data of the programming job with the correct answer template set, and output corresponding confidence data;

S400. Confirm the correction result of the programming job according to the confidence data.

In this embodiment, when the user performs programming learning on, for example, an online programming education platform, he inputs the programming assignment to be corrected in the corresponding code editing area according to the job requirements, and the online programming education platform can automatically obtain the pending correction submitted by the user. Specifically, the programming job to be corrected may be a programming job submitted in history to be corrected, or a programming job submitted in real time to be corrected, which is not limited in this embodiment. In particular, in this embodiment, it can be The programming operation in the form of graphical programming is described, that is, the programming operation contains a plurality of encapsulated building block code blocks with corresponding functions. Of course, in other embodiments, the programming operation can also be in other forms such as text, etc. This embodiment does not limit this.

After the programming job to be corrected is obtained, feature extraction is performed through a neural network model that is pre-built and trained, and the feature data of the programming job to be corrected is calculated and output. The neural network model can be a VGG neural network, which has a simple structure. , wide feature width, and the ability to extract more feature information. Of course, in other embodiments, neural network models with other structures can also be used, which are not limited in this embodiment; the specific training process of the neural network model is based on pre- The collected correction training samples are used to train the neural network model, and the model training is completed when the neural network model reaches the preset convergence condition. The programming operation, input it into the neural network model, and perform feature extraction on the corrected training samples according to the neural network model, and reverse the regulator weight parameters until the model converges, that is to say, the trained neural network model can accurately Identify and extract the feature data corresponding to the building block code blocks contained in the programming job, and provide a reliable data basis for subsequent automatic corrections.

After extracting the feature data of the programming job to be corrected, perform feature matching with the correct answer template set, and calculate and output the corresponding confidence data. For example, at least one correct answer template can be preset at the beginning of automatic correction and extracted through the neural network model The feature data is stored in the correct answer template set, and in the subsequent automatic correction, the feature data of the programming job to be corrected is matched with the template feature data contained in the correct answer template set, and further confirmed according to the confidence data output by the calculation The correction result of the programming job to be corrected is obtained. In this embodiment, the neural network model and the feature matching process are obtained by combining deep learning training, and the abstract graphical comparison during automatic correction of the programming job is converted into accurate feature data matching, effectively improving The efficiency of automatic correction of homework in the process of graphical programming teaching is improved.

In one embodiment, performing feature matching between the feature data of the programming job and the correct answer template set, and outputting corresponding confidence data, including:

Input the feature data of the programming job into a preset matching model, and output the confidence data between the programming job and the preset template category, the matching model is based on all the obtained correct answer template sets. The correct answer template is trained.

In this embodiment, during feature matching, a preset matching model is used to classify and match the feature data of the programming job, and the matching model is obtained by training based on the obtained correct answer templates in the correct answer template set , that is, a matching model is obtained by pre-training the obtained correct answer template. The matching model is used to perform feature matching between the input programming job and the preset template category and output the corresponding confidence data. The confidence data represents the programming job. The similarity between the feature data of the preset template category and the feature data corresponding to the preset template category, the sum of the confidence data of the same programming job for all preset template categories is 1, that is, the matching model can match the output programming job and each preset template category. Similarity between template categories to achieve efficient feature matching processing for input feature data of programming jobs.

In one embodiment, before the feature data of the programming job is input into the preset matching model, the method further includes:

Classify all the correct answer templates that have been obtained in the correct answer template set, and mark each correct answer template as a corresponding preset template category;

Input each correct answer template into the neural network model respectively, and calculate and output the characteristic training data of each correct answer template;

The matching model is constructed, and the matching model is trained according to the preset template category of each correct answer template and the feature training data, until the matching model satisfies preset convergence conditions.

In this embodiment, before feature matching is performed, a matching model for classification matching is constructed and trained, firstly, all correct answer templates obtained in the correct answer template set are classified into categories, and each correct answer template is marked as corresponding The preset template category, and the feature training data of each correct answer template is calculated by the preset neural network model, so as to constitute the classification training sample of the matching model, that is, the classification training sample of the matching model is the feature training that has been classified into categories. The matching model is trained through the classification training sample, and the model training is completed when the matching model satisfies the preset convergence adjustment. The matching model after model training can accurately classify and match the characteristic data of the programming job to obtain the Confidence data between jobs and each preset template category to achieve efficient and accurate end-to-end feature matching.

Optionally, since the correct answer template set may be dynamically updated according to the batch improvement process to meet the correction needs of various implementation forms, the matching model training and classification matching can be performed at the same time. After the model training of the matching model is completed for the first time , you can continue to train according to the feature training data contained in the dynamically updated correct answer template set and the preset template category to obtain an updated matching model, improve the template coverage rate when performing feature matching on the feature data of the programming job, and ensure that the feature When matching, there will be no correction errors due to omissions, which improves the reliability of automatic corrections.

The confirming the correction result of the programming job to be corrected according to the confidence data includes:

If there is confidence data greater than or equal to the confidence threshold, the correction result of the programming job is qualified;

If the confidence level data are all less than the confidence level threshold, the correction result of the programming job is unqualified.

In this embodiment, after feature matching is performed to obtain confidence data corresponding to each preset template category, if there is confidence data greater than or equal to the confidence threshold, it means that the current programming job to be corrected corresponds to the confidence data. The preset template categories are matched, and the correction result of the programming job is confirmed to be qualified, so the correction result of the programming job is confirmed to be qualified. The specific confidence threshold can be flexibly set according to the actual correction requirements and the coverage rate of the correct answer template set. , this embodiment does not limit this; if it does not exist, that is, the confidence data are all smaller than the confidence threshold, it means that there is no template category matching the programming job to be corrected in the current preset template category, and the programming job is confirmed at this time. The correction result is unqualified, and it is marked as unknown work for subsequent processing. By adopting the confidence threshold and the confidence threshold can be adjusted, the coverage rate of the correct answer template can be flexibly adjusted according to the correction needs, so that it can be judged without exact matching. Whether the programming job is correct effectively improves the efficiency of automatic correction of programming jobs.

In one embodiment, if the confidence level data are all smaller than the confidence level threshold, after the correction result of the programming job is unqualified, the method may further include:

classifying the programming jobs whose correction result is unqualified into the set of jobs to be processed;

Cluster processing is performed on the set of jobs to be processed, new correct answer templates are extracted and classified into the set of correct answer templates, and feature matching is continued until all correct answers are matched from all programming jobs to be corrected.

In this embodiment, if the programming job currently to be corrected is marked as an unknown work after feature matching, it is classified into the job set to be processed, and further clustering processing is performed on the job set to be processed. A new correct answer template is extracted from the processing homework set and merged into the correct answer template set, that is, although the correct answer of the same homework can have multiple implementation combinations, the programming homework to be corrected during the initial matching cannot be compared with the existing one. The set of correct answer templates is successfully matched, but the total number of correct answers is theoretically determined, and the correct answer assignments have commonalities, so there is a certain logical connection between different correct answer templates, such as the key execution in different correct answer templates. The steps are the same, and the remaining execution steps are reversed in order, etc. Therefore, in this embodiment, after feature matching, according to the matching results, the unknown works are clustered, and a new correct answer template is gradually obtained, so as to perform the correct answer template set. Update, further continue feature matching through the updated correct answer template set, and repeat the above matching, set entry and clustering process to form a virtuous circle of matching and clustering, until all correct answer templates are obtained from the clustering. All the correct answers are matched in all the programming jobs to be corrected, which effectively improves the coverage and efficiency of automatic correction of programming jobs.

The clustering process is performed on the set of jobs to be processed, and a new correct answer template is extracted and classified into the correct answer template set, and then feature matching is continued, including:

Perform clustering processing on the set of jobs to be processed according to a preset clustering algorithm;

Determine whether there is a pending job category that meets the preset clustering criteria;

If so, extract a new correct answer template from the to-be-processed job category that satisfies the preset clustering criteria and classify it into the correct answer template set;

Continue to perform assignment matching on programming assignments to be marked through the updated set of correct answer templates.

In this embodiment, the clustering operation is performed on the set of works to be processed that cannot be successfully matched temporarily in the matching process, and a preset clustering algorithm such as K-means (K-means) clustering, mean-shift clustering, and density-based clustering is used. The clustering algorithm such as the clustering algorithm performs clustering processing on the set of jobs to be processed. Since the correct answer jobs have commonalities, preset clustering criteria can be set according to the requirements of the current job, and the jobs to be processed can be judged during clustering processing. Whether there is a pending job category that meets the preset clustering criteria, if so, extract the works closest to the preset clustering criteria as a new correct answer template, and put it into the correct answer template set. The updated set of correct answer templates continues to perform assignment matching on programming assignments to be marked. Specifically, the existing correct answer template can be used as the initial clustering center, and for each to-be-processed job sample in the to-be-processed job set, the distance between each to-be-processed job sample and each initial clustering center is calculated respectively. When the distance between the job sample to be processed and any initial cluster center is less than the preset distance, it is deemed to meet the preset clustering standard, and the job sample to be processed that meets the preset clustering standard is assigned to the initial clustering center with the smallest distance. In the corresponding to-be-processed job categories, the to-be-processed job samples closest to the initial cluster center are extracted from each to-be-processed job category as a new correct answer template, that is, the to-be-processed job category corresponds to the already obtained correct answer template, It represents an assignment category similar to each correct answer template, and further extracts the most similar work to the corresponding correct answer template from each pending assignment category as a new correct answer template to update the correct answer template set to improve Coverage of the set of correct answer templates.

That is to say, in this embodiment, after the feature matching is performed, a new correct answer template is obtained by performing a clustering operation on the set of jobs to be processed, so as to update the correct answer template set, so as to increase the correct answer template set when correcting and matching. Coverage, after repeating this process, you can gradually match the correct answers from all the programming assignments to be corrected, and gradually improve the correct answer template set corresponding to the current assignment, effectively improving the coverage and efficiency of subsequent assignments.

The extracting a new correct answer template from the to-be-processed job category that meets the preset clustering criteria and classifying it into the correct answer template set specifically includes:

The cluster centers are extracted for the to-be-processed job categories that meet the preset clustering criteria, and the extracted cluster centers are used as a new correct answer template and included in the correct answer template set.

In this embodiment, when extracting a new correct answer template, the cluster centers are taken for the job categories to be processed that meet the preset clustering criteria, and the extracted cluster centers are used as the new correct answer template, that is, according to each After the distance between the job samples to be processed and each initial cluster center is divided into categories according to the principle of minimum distance, the to-be-processed job category corresponding to each initial cluster center is obtained. The job sample recalculates the cluster center, calculates the average distance between the to-be-processed job sample and the initial cluster center in the to-be-processed job category, and uses the to-be-processed job sample with the distance closest to the average value as the cluster of the to-be-processed job category As the cluster center, the pending job samples are the works that are closest to the correct answer template in the corresponding pending assignment category, ensuring that each time a new correct answer template is extracted from the pending assignment collection, the closest preset cluster is selected. Class criteria, the assignments that best meet the correct answer requirements, ensure the accuracy of the correct answer template collection.

Optionally, every time a new correct answer template is extracted from the set of jobs to be processed, the new correct answer template is removed from the set of jobs to be processed to avoid repeated extraction during subsequent clustering algorithms. , causing other correct assignments to fail to be successfully extracted and affecting the coverage of the correct answer template set.

Optionally, before the inputting the programming job into a preset neural network model, before calculating and outputting the characteristic data of the programming job, the method may further include:

The programming job is serialized according to preset rules.

In this embodiment, since the work of graphical programming is a building block combination formed by splicing multiple building block code blocks, it is impossible to automatically identify the code data represented by each building block code block combination. Improve the efficiency of subsequent matching and clustering. Before performing job matching, the programming job to be corrected is vectorized, and the graphical programming job is converted into an accurate digital programming job, so as to realize the automatic programming job. Recognition provides a reliable data basis for efficient and accurate automatic correction.

Optionally, performing serialization processing on the programming job according to preset rules includes:

Get the building block code block combination in the programming assignment;

Analyze the building block code block combination in a preset order to obtain the building block type and building block data corresponding to each building block code block;

A corresponding serialized array is generated after serializing according to the building block type and building block data according to preset rules.

In this embodiment, the specific serialization process is to first obtain the building block code block combination in the programming job to be corrected. As shown in FIG. 2 , the programming job to be corrected is formed by superimposing multiple building block code blocks. Combination of building block code blocks, each building block code block is encapsulated with codes that implement corresponding functions, such as JavaScript code, etc. After obtaining the building block code block combination, it is parsed in a preset order, because the obtained building block code block There is a certain timing relationship between different building block code blocks in the combination, so the combination of building block code blocks is parsed in a preset order, and the building block type and building block data corresponding to each building block code block arranged in the preset order are obtained to ensure that The function and data represented by each building block code block are obtained without affecting the logical relationship between the building block code blocks, and then the block type and building block data obtained by parsing are serialized according to preset rules to generate The serialized array corresponding to the code block combination enables the graphical block code block combination to be converted into an easily identifiable quantized array form during automatic correction of programming operations. Recognition efficiency and accuracy.

Specifically, the building block type may be one or more of constant building blocks, variable building blocks and/or operation building blocks, that is, the obtained building block code block combination may include event building blocks, control building blocks, constant building blocks, variable building blocks, and operation building blocks One or more of the building blocks, through the combination of a variety of different types of building block code blocks to achieve corresponding functions, wherein the operation building blocks may further include, for example, function building blocks, operation building blocks, appearance building blocks, sound building blocks, etc. It needs to be set, which is not limited in this embodiment. The building block data is the input parameters on the building block code block. For example, taking the control building block as an example, if the current control building block is "wait for 0.5 seconds", the building block data is "0.5". Of course, there is no input parameter in the building block code block. , the block data can be empty, such as the "repeatedly executed" control block and so on.

Optionally, the works included in the initial correct answer template set are also serialized according to preset rules and feature extraction of neural network models, so that intuitive and accurate quantitative matching can be directly performed during subsequent job matching, effectively improving matching efficiency and accuracy.

Optionally, parsing the building block code block combination in a preset order to obtain the building block type and building block data corresponding to each building block code block, including:

Obtain the splicing sequence of the building block code block combination;

Each building block code block is parsed in sequence according to the splicing sequence, and the building block types and building block data arranged according to the splicing sequence are obtained.

In this embodiment, the building block code block combination in the programming job to be corrected is input by the user, and the specific user can directly select the packaged building block code block in the preset building block library, and drag and drop in the code editing area. Edit operations of building block code such as dragging, splicing, and combining, and then obtain the building block code block combination with the corresponding splicing sequence. In the graphics programming scenario, all building blocks code blocks are placed in sequence, so there is a natural sequence relationship in the time sequence. By obtaining the splicing sequence of the building block code block combination, the stacking sequence when the user edits the input can be obtained. , and then analyze each building block code block in turn according to the splicing sequence, and obtain the building block type and building block data arranged according to the splicing sequence, so as to parse and restore the functions and data of the stacked building block code blocks in order to ensure the subsequent sequence. The function and data corresponding to the building block code block will not have logical timing errors during the processing, and the accuracy of the building block code block recognition will be improved.

Optionally, generating a corresponding serialized array after serializing according to the building block type and building block data according to preset rules, including:

The building block type and building block data are respectively converted into corresponding encoded data according to a preset encoding mapping library, and the encoding mapping library is used to store the mapping relationship between the building block type and the building block data and the encoded data;

After encapsulating the encoded data according to the preset data structure, a one-dimensional array corresponding to each building block code block is obtained;

The one-dimensional arrays are superimposed according to the splicing sequence to generate a two-dimensional serialized array corresponding to the combination of the building block code blocks.

In this embodiment, during the serialization process, all building block types and building block data arranged according to the splicing sequence are encoded in sequence and then converted into corresponding encoded data respectively. Specifically, a one-to-one correspondence can be performed according to a preset encoding mapping library. Data conversion, the coding mapping library stores the mapping relationship between the building block type and the coding data, and the mapping relationship between the building block data and the coding data, and searches the coding mapping library for corresponding After the building block type and building block data, the corresponding encoded data can be obtained to realize convenient and fast data conversion, and then the digitized encoded data is encapsulated according to the preset data structure to obtain a one-dimensional array corresponding to each building block code, that is In this embodiment, each building block code block is correspondingly converted into a one-dimensional array, and after the one-dimensional array corresponding to each building block code is obtained, all the one-dimensional arrays are superimposed according to the splicing sequence to generate a two-dimensional sequence The two-dimensional serialized array corresponds to the combination of building block code blocks, which has the same splicing sequence and the encoded data corresponding to the building block type and building block data one-to-one. Through the two-dimensional serialized array, the graphic The combined combination of building block code blocks is converted into quantified two-dimensional encoded data, which realizes the vectorized processing of graphical programming works, facilitates the subsequent realization of accurate building block code identification and review, and is conducive to the automatic review of graphical programming works.

Optionally, before generating a corresponding serialized array after serializing according to the building block type and building block data according to a preset rule, the method further includes:

The encoding mapping library is obtained by predefining type encoding data corresponding to all building block types and action encoding data corresponding to the building block data.

In this embodiment, the building block types and building block data included in all building block code blocks in the building block library are encoded, defined and stored in advance, and the encoding mapping library is obtained to realize accurate data conversion. Specifically, the encoding mapping library can use a mapping table As shown in Table 1, it shows the type coding data corresponding to some building block types, so that during subsequent serialization processing, the type coding data corresponding to each building block type can be directly searched and called, and the action coding data corresponding to the building block data is Yes, to ensure the consistency and accuracy of each serialization process.

Table 1

Optionally, the first data bit of the one-dimensional array is type-encoded data, and the remaining data bits are action-encoded data and/or preset space characters.

In this embodiment, the converted encoded data is encapsulated according to a preset data structure, so that each block code to be processed representing a variable, constant or operation is quantified into a serialized one-dimensional array in a uniform data format. The data structure As shown in Table 2, the length of each line of data can be fixed to N bits, and N can be based on the actual building block, as long as it can meet the longest demand. The first bit is the type encoding data, and the last few bits are the action encoding data. , can be replaced with a preset blank character such as 0.

Table 2

Please refer to FIG. 2 for details. The serialization process in the automatic correction method for programming jobs provided by the present invention will be described with specific application examples. As shown in FIG. 2 , in this embodiment, the combination of building blocks and code blocks is repeated according to the splicing sequence. Execute, next action, wait for 0.5 seconds, the corresponding block types are repeated execution, action, wait (in seconds), and the corresponding block data are no data, action code data corresponding to the next action, 0.5, so After encapsulating the coding data corresponding to the building block type and building block data according to the above preset data structure, the building block code block can be serialized and converted into a two-dimensional serialized array as shown in Table 3, so as to realize the conversion from graphic to digital. Achieve the purpose of automatic identification of graphic programming works, and provide accurate identification results for automatic correction of graphic programming works.

table 3

Specifically, in order to better understand the implementation process of the method for automatic correction of programming jobs provided by the present invention, the following is a description of the correction process of programming jobs with specific application examples in conjunction with FIG. 3 and FIG. 4 :

In this embodiment, the automatic correction of programming jobs is divided into two processes. First, the feature calculation process as shown in Figure 3 is performed. After the jobs to be corrected are obtained, they are serialized first, so that the graphic programming works can be converted into the ones that can be used for calculation. vector data, and then input the serialized data into the pre-trained VGG neural network for calculation, output the feature data of the job to be corrected, and then perform the template matching process shown in Figure 4, which is a step-by-step iterative recursion The process of first obtaining the feature data of the job to be corrected that was calculated and output before, and then performing the feature matching calculation between the feature data of the job to be corrected and the existing correct answer template set, and obtaining several confidence data with the same number of correct answer templates. , the correct answer template initially included in the correct answer template set also undergoes the same serialization processing and feature extraction to achieve accurate matching calculation, and then judge whether there is data greater than or equal to the confidence threshold in several confidence data, if so , then mark the work to be corrected as the correct answer and put it into the correct answer template set, if not, it will be classified into the unknown result work set; further use K-means and other clustering algorithms to cluster the unknown result work set, judge Whether there are works that meet the clustering criteria in the results, if so, select the cluster center as the new correct answer template in the set of works that meet the clustering criteria, and classify the new template that has been successfully clustered into the correct answer template set , and remove it from the unknown result work set, continue to match the homework through the updated correct answer template set, repeat the above matching, set, and clustering process, you can gradually match and detect the correct homework from all homework, and Gradually improve the correct answer template of the current homework, effectively improving the coverage and efficiency of matching and correcting subsequent works.

It can be seen from the above method embodiments that the automatic correction method for programming jobs provided by the present invention calculates the feature data of programming jobs through a neural network model, and performs feature matching with the correct answer template to obtain corresponding confidence data, which is combined with deep learning training. The confidence data obtained by matching with the feature can effectively and automatically correct the programming assignments, which improves the efficiency of assignment correction in the programming teaching process.

It should be noted that the above steps do not necessarily have a certain sequence. Those of ordinary skill in the art can understand from the description of the embodiments of the present invention that in different embodiments, the above steps may have different execution orders. That is, it can be executed in parallel, or it can be executed interchangeably, and so on.

Another embodiment of the present invention provides an apparatus for automatic correction of programming operations. As shown in FIG. 5 , the apparatus 1 includes:

an acquisition module 11, used for acquiring the programming job to be corrected;

A feature calculation module 12, configured to input the programming job into a preset neural network model, and calculate and output the feature data of the programming job;

The matching module 13 is used to perform feature matching between the feature data of the programming job and the correct answer template set, and output corresponding confidence data;

The output module 14 is configured to confirm the correction result of the programming job according to the confidence data.

The acquisition module 11 , the feature calculation module 12 , the matching module 13 and the output module 14 are connected in sequence. For specific implementations, please refer to the above corresponding method embodiments, which will not be repeated here.

Further, the feature calculation module 12 is specifically used for:

The feature data of the programming job is put into a preset matching model, and the confidence data between the programming job and the preset template category is output, and the matching model is based on all the correct answers obtained in the correct answer template set. The answer template is trained.

Further, the programming job automatic correction device also includes:

a labeling module, configured to label all the correct answer templates obtained in the correct answer template set, and label each correct answer template as a corresponding preset template category;

an input module for inputting each correct answer template into the neural network model respectively, and calculating and outputting the characteristic training data of each correct answer template;

A training module, configured to construct the matching model, and train the matching model according to the preset template category of each correct answer template and the characteristic training data, until the matching model satisfies preset convergence conditions.

Further, the output module 14 is used for:

If there is confidence data greater than or equal to the confidence threshold, output the correction result of the programming job as qualified;

If the confidence data are all smaller than the confidence threshold, output the correction result of the programming job as unqualified.

Further, the output module 14 also includes:

a categorization unit, used for categorizing the programming jobs marked as unknown works whose correction result is unqualified into the set of jobs to be processed;

The clustering unit is used to perform clustering processing on the set of tasks to be processed, extract new correct answer templates and classify them into the set of correct answer templates and continue feature matching until all programming tasks to be corrected are matched. correct answer.

Further, the programming job automatic correction device also includes:

The serialization module is used to serialize the programming job according to preset rules.

Further, the serialization module includes:

Get unit, used to get the building block code block combination in the programming job;

a parsing unit for parsing the building block code block combination in a preset order to obtain the building block type and building block data corresponding to each building block code block;

The serialization unit is used to generate a corresponding serialized array after serialization according to the preset rules according to the building block type and the building block data.

Further, the parsing unit includes:

a timing acquisition subunit for acquiring the splicing timing of the building block code block combination;

The parsing subunit is used to parse each building block code block in sequence according to the splicing sequence to obtain the building block types and building block data arranged according to the splicing sequence.

Further, the serialization unit includes:

The conversion subunit is used to convert the building block type and building block data into corresponding encoded data according to a preset encoding mapping library, and the encoding mapping library is used to store the building block type and building block data and the encoding data. The mapping relationship between;

A data encapsulation subunit for encapsulating the encoded data according to a preset data structure to obtain a one-dimensional array corresponding to each building block code block;

The sequence stacking subunit is used for stacking the one-dimensional arrays according to the splicing sequence to generate a two-dimensional serialized array corresponding to the combination of the building block code blocks.

Another embodiment of the present invention provides an automatic correction system for programming jobs. As shown in FIG. 6 , the system 10 includes:

One or more processors 110 and the memory 120 are described by taking one processor 110 as an example in FIG. 6 . The processor 110 and the memory 120 may be connected by a bus or in other ways, and the connection by a bus is used as an example in FIG. 6 .

The processor 110 is used to complete various control logics of the system 10, and it can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a single-chip microcomputer, an ARM (Acorn RISCMachine) or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination of these components. Also, the processor 110 may be any conventional processor, microprocessor or state machine. The processor 110 may also be implemented as a combination of computing devices, eg, a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in combination with a DSP, and/or any other such configuration.

The memory 120, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as those corresponding to the method for automatically correcting programming jobs in the embodiments of the present invention. program instructions. The processor 110 executes various functional applications and data processing of the system 10 by running the non-volatile software programs, instructions and units stored in the memory 120, that is, to implement the automatic correction method of programming jobs in the above method embodiments.

The memory 120 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the system 10 and the like. Additionally, memory 120 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, memory 120 may optionally include memory located remotely from processor 110, which may be connected to system 10 through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

One or more units are stored in the memory 120, and when executed by one or more processors 110, execute the automatic correction method for programming jobs in any of the above method embodiments, for example, execute the method step S100 in FIG. 1 described above Go to step S400.

Embodiments of the present invention provide a non-volatile computer-readable storage medium, where the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by one or more processors, for example, to execute the above-described The method steps S100 to S400 in FIG. 1 .

As examples, non-volatile storage media can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) as external cache memory. By way of illustration and not limitation, RAM may be configured in a variety of formats such as Synchronous RAM (SRAM), Dynamic RAM, (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus (Lambas) RAM (DRRAM) and many other forms. The disclosed memory components or memories of the operating environments described herein are intended to include one or more of these and/or any other suitable types of memory.

To sum up, in the method, system and medium for automatic correction of programming jobs disclosed in the present invention, the method obtains the programming job to be corrected; inputs the programming job into a preset neural network model, and calculates and outputs the feature data of the programming job; feature matching the feature data of the programming job with the correct answer template set, and output corresponding confidence data; confirm the correction result of the programming job according to the confidence data. In the embodiment of the present invention, the feature data of the programming job is calculated by the neural network model, and the corresponding confidence data is obtained after feature matching between the feature data and the correct answer template. Efficient automatic correction improves the efficiency of homework correction in the process of programming teaching.

Of course, those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware (such as processors, controllers, etc.) through a computer program, and the computer program can be stored in In a non-volatile computer-readable storage medium, the computer program may include the processes of the above-mentioned method embodiments when executed. The storage medium may be a memory, a magnetic disk, a floppy disk, a flash memory, an optical memory, and the like.

It should be understood that the application of the present invention is not limited to the above examples. For those of ordinary skill in the art, improvements or transformations can be made according to the above descriptions, and all these improvements and transformations should belong to the protection scope of the appended claims of the present invention.

Claims (10)

1. An automatic correcting method for programming operation is characterized by comprising the following steps:

acquiring a programming job to be corrected;

inputting the programming operation into a preset neural network model, and calculating and outputting characteristic data of the programming operation;

performing feature matching on the feature data and a correct answer template set, and outputting corresponding confidence coefficient data;

and confirming the correction result of the programming job according to the confidence coefficient data.

2. The method according to claim 1, wherein the performing feature matching on the feature data of the programming job and the correct answer template set and outputting corresponding confidence level data comprises:

inputting the feature data of the programming operation into a preset matching model, and outputting confidence coefficient data between the programming operation and a preset template category, wherein the matching model is obtained by training based on all correct answer templates obtained from the correct answer template set.

3. The method according to claim 2, wherein before inputting the characteristic data of the programming job into the preset matching model, the method further comprises:

performing category marking on all correct answer templates obtained in the correct answer template set, and respectively marking each correct answer template as a corresponding preset template category;

inputting each correct answer template into the neural network model respectively, and calculating and outputting feature training data of each correct answer template;

and constructing the matching model, and training the matching model according to the preset template type and the feature training data of each correct answer template until the matching model meets a preset convergence condition.

4. The method of claim 2, wherein the validating the approval of the programming job based on the confidence data comprises:

if the confidence coefficient data which is greater than or equal to the confidence coefficient threshold exists, the correction result of the programming operation is qualified;

if the confidence coefficient data are all smaller than the confidence coefficient threshold value, the correction result of the programming operation is unqualified;

the programming jobs with the modification results being unqualified are classified into a job set to be processed;

and clustering the job set to be processed, extracting a new correct answer template, classifying the new correct answer template into the correct answer template set, and continuing to perform feature matching until all correct answers are matched from all the programming jobs to be corrected.

5. The method according to claim 1, wherein before inputting the programming job into a preset neural network model and calculating and outputting the characteristic data of the programming job, the method further comprises:

and carrying out serialization processing on the programming operation according to a preset rule.

6. The automatic correction method for programming jobs according to claim 5, wherein the serializing the programming jobs according to preset rules comprises:

acquiring a building block code block combination in the programming operation;

analyzing the building block code block combination according to a preset sequence to obtain a building block type and building block data corresponding to each building block code block;

and serializing according to the building block type and the building block data according to a preset rule to generate a corresponding serialized array.

7. The automatic correcting method for programming operation according to claim 6, wherein the analyzing the combination of the block code blocks according to the preset sequence to obtain the block type and the block data corresponding to each block code block comprises:

acquiring a splicing time sequence of the building block code block combination;

and analyzing each building block code block in sequence according to the splicing time sequence to obtain the building block types and building block data arranged according to the splicing time sequence.

8. The method of claim 7, wherein the generating a corresponding serialized array after serializing according to the building block type and building block data according to a predetermined rule comprises:

respectively converting the building block type and the building block data into corresponding coded data according to a preset coding mapping library, wherein the coding mapping library is used for storing the mapping relation between the building block type, the building block data and the coded data;

packaging the coded data according to a preset data structure to obtain a one-dimensional array corresponding to each building block code block;

and superposing the one-dimensional arrays according to a splicing time sequence to generate a two-dimensional serialized array corresponding to the building block code block combination.

9. An automatic program job modification system, comprising at least one processor; and the number of the first and second groups,

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method for automatic program job modification according to any one of claims 1-8.

10. A non-transitory computer-readable storage medium storing computer-executable instructions that, when executed by one or more processors, cause the one or more processors to perform the method for automatic program job modification according to any one of claims 1-8.

Images