Test item file screening method and model training method, device, equipment and medium

Abstract

The present disclosure provides a test question file screening method and device, a model training method and device, and a medium. The present disclosure relates to the technical field of artificial intelligence, specifically to the technical field of target recognition, and can be applied to test question file screening and the like. The specific implementation scheme is as follows: feature information of multiple dimensions of a target file to be recognized is extracted, the dimensions of the target file including a title dimension, an abstract dimension, a text dimension, and a file format dimension; whether the target file belongs to a test question file is determined based on the feature information of the multiple dimensions; when the target file is a test question file, at least one preset test question label information is extracted from the target file, and the target file is classified based on the test question label information. The above process can improve the accuracy of the recognition result of the file, the screening process of the test question file can be automatically completed, the efficiency of test question screening is improved, and the labor cost is reduced.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, specifically to the field of target recognition technology, and can be applied to scenarios such as test question file screening. Background Technology

[0002] In building a question bank for question-searching products, it's necessary to sift through a large number of files to extract question files, then break these files down into multiple individual questions for display to users during their search. Current methods for filtering question files typically involve initially identifying potential question files based on limited information, then manually verifying their authenticity by reviewing the text content, and manually adding tags to categorize the files. However, this initial screening process is prone to misjudging and missing some genuine question files. Furthermore, manual verification is costly and inefficient. Summary of the Invention

[0003] This disclosure provides a method for screening test question files, a method for training models, an apparatus, equipment, and a medium.

[0004] According to a first aspect of this disclosure, a method for screening test question files is provided, comprising:

[0005] Extract feature information from multiple dimensions of the target file to be identified. The dimensions of the target file include title dimension, abstract dimension, text dimension, and file format dimension.

[0006] The identification result determines whether the target file belongs to the test question file based on feature information from multiple dimensions;

[0007] When the target file is a test question file, at least one preset test question tag information is extracted from the target file, and the target file is classified based on the test question tag information.

[0008] According to a second aspect of this disclosure, a method for training a test item recognition model is provided, comprising:

[0009] Label each sample file in multiple sample files to determine whether it belongs to the actual result of the test question file;

[0010] Extract feature information from multiple dimensions of each sample file from multiple sample files. The dimensions of the sample files include title dimension, abstract dimension, body text dimension, and file format dimension.

[0011] Multi-dimensional feature information is input into the test item recognition model, and the test item recognition model is used to output the recognition result of whether the sample file belongs to the test item file.

[0012] Based on the actual results and recognition results of each sample file, the recognition accuracy of the test item recognition model is determined;

[0013] The parameters of the test item recognition model are adjusted based on the recognition accuracy and the preset accuracy threshold.

[0014] According to a third aspect of this disclosure, a test item file screening device is provided, comprising:

[0015] The target feature extraction module is used to extract feature information of the target file to be identified from multiple dimensions, including title dimension, abstract dimension, body text dimension and file format dimension;

[0016] The target result output module is used to determine whether the target file belongs to the test question file based on feature information from multiple dimensions.

[0017] The test question classification module is used to extract at least one preset test question tag information from the target file when the target file is a test question file, and classify the target file based on the test question tag information.

[0018] According to a third aspect of this disclosure, a test item recognition model training apparatus is provided, comprising:

[0019] The results annotation module is used to annotate whether each sample file in multiple sample files belongs to the actual results of the test question file;

[0020] The sample feature extraction module is used to extract feature information from each of the multiple sample files in multiple dimensions. The dimensions of the sample files include title dimension, abstract dimension, body text dimension and file format dimension.

[0021] The sample result output module is used to input multi-dimensional feature information into the test item recognition model and use the test item recognition model to output the recognition result of whether the sample file belongs to the test item file;

[0022] The accuracy determination module is used to determine the recognition accuracy of the test item recognition model based on the actual results and recognition results of each sample file.

[0023] The parameter adjustment module is used to adjust the parameters of the test item recognition model based on the recognition accuracy and a preset accuracy threshold.

[0024] According to a fifth aspect of this disclosure, an electronic device is provided, comprising:

[0025] At least one processor; and

[0026] A memory communicatively connected 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 test item file screening method or test item recognition model training method.

[0027] According to a sixth aspect of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to perform the above-described test item file screening method or test item recognition model training method.

[0028] According to the seventh aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described test question file screening method or test question recognition model training method.

[0029] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description.

[0030] The beneficial effects of the technical solution provided in this disclosure are:

[0031] In this disclosed technical solution, information from multiple dimensions, including the document's title, abstract, body, and file format, is used to identify whether a document belongs to the test question category. For target documents that do belong to the test question category, classification can be automatically completed based on their contained tag information. The above process can significantly improve the accuracy of document identification results, largely avoiding the omission of genuine test question documents due to misjudgment. The test question document screening process can be completed automatically, which not only improves the efficiency of test question screening but also reduces manual costs. Attached Figure Description

[0032] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0033] Figure 1 A flowchart illustrating a test question file screening method provided in an embodiment of this disclosure is shown;

[0034] Figure 2 A flowchart illustrating another test question file screening method provided in an embodiment of this disclosure is shown;

[0035] Figure 3 A flowchart illustrating a test item recognition model training method provided in an embodiment of this disclosure is shown.

[0036] Figure 4 This illustration shows one of the schematic diagrams of a test question file screening device provided in an embodiment of the present disclosure;

[0037] Figure 5 A second schematic diagram of a test question document screening device provided in an embodiment of this disclosure is shown;

[0038] Figure 6 A schematic block diagram of an example electronic device is shown that can be used to implement the test question file screening method provided in the embodiments of this disclosure. Detailed Implementation

[0039] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0040] In building a question bank for question-searching products, it's necessary to sift through a large number of files to extract question files, then break these files down into multiple individual questions for display to users during their search. Current methods for filtering question files typically involve initially identifying potential question files based on limited information, then manually verifying their authenticity by reviewing the text content, and manually adding tags to categorize the files. However, this initial screening process is prone to misjudging and missing some genuine question files. Furthermore, manual verification is costly and inefficient.

[0041] The test question file screening method, model training method, apparatus, device, and medium provided in this disclosure are intended to solve at least one of the above-mentioned technical problems in the prior art.

[0042] Figure 1 This illustration shows a flowchart of a test question file screening method provided in an embodiment of the present disclosure, such as... Figure 1 As shown, the method mainly includes the following steps:

[0043] S110: Extract feature information from multiple dimensions of the target file to be identified.

[0044] In this disclosure, the dimensions of the target document include the title dimension, abstract dimension, body dimension, and document format dimension, and may also include other dimensions. The type of feature information corresponding to each dimension can be preset. This feature information may reflect whether the target document is a test document to a certain extent. For example, the feature information may be keywords, high-frequency words, specified sentence structures and format types, etc.

[0045] S120: The identification result determines whether the target file belongs to the test question file based on feature information from multiple dimensions.

[0046] In this embodiment of the disclosure, feature information of multiple dimensions can be input into a pre-trained test item recognition model, and the test item recognition model can output a recognition result of whether the target file belongs to a test item file. Optionally, in this embodiment of the disclosure, when using the test item recognition model to output a recognition result of whether the target file belongs to a test item file, the test item recognition model can determine the credibility of the target file belonging to a test item file based on the feature information of each dimension; and determine the recognition result of whether the target file belongs to a test item file based on the credibility corresponding to the feature information of multiple dimensions.

[0047] The embodiments disclosed herein can train an initial test item recognition model. For example, for multiple sample files, each sample file in the multiple sample files can be pre-labeled to indicate whether it belongs to the actual result of the test item file. Then, the multiple sample files are used as a training set and input into the initial test item recognition model for training. The specific training process of the test item recognition model will be described in detail in the following content.

[0048] After inputting multi-dimensional feature information of the target file into the pre-trained test item recognition model, the test item recognition model can output a recognition result, which indicates whether the target file belongs to the test item file. When the recognition result indicates that the target file belongs to the test item file, step S130 can be executed; when the recognition result indicates that the target file does not belong to the test item file, the process of the method can be terminated, or step S110 can be re-executed for the next target file.

[0049] In this embodiment of the disclosure, it can be determined whether the feature information of each dimension of the target file belongs to the preset test question feature information. If the feature information of one dimension of the target file belongs to the preset test question feature information, then the feature information can be determined as the target feature information. When the number of target feature information of the target file is greater than a preset value, it can be determined that the target file belongs to the test question file.

[0050] S130: When the target file is a test question file, extract at least one preset test question tag information from the target file, and classify the target file based on the test question tag information.

[0051] It is understood that test question files typically contain information that characterizes their type; in this embodiment, such information is defined as test question tag information. When a target file is determined to be a test question file, the test question tag information can be extracted from it, and the test question file type can be determined based on the tag information, thus facilitating the categorization and management of the selected target files by type.

[0052] The test question file screening method provided in this disclosure comprehensively considers multiple dimensions of information, including the file's title, abstract, body text, and file format, to identify whether a file belongs to the test question file category. For target files that do belong to the test question file category, classification can be automatically completed based on their contained tag information. The above process can significantly improve the accuracy of file identification results, largely avoiding the omission of genuine test question files due to misjudgment. The test question file screening process can be completed automatically, which not only improves the efficiency of test question screening but also reduces manual costs.

[0053] Optionally, the type and quantity of test question tag information extracted from the target file in this embodiment of the present disclosure can be determined according to a preset rule. For example, at least one of the regional information, school level information and subject information can be extracted from the target file as test question tags.

[0054] It should be noted that, in this embodiment of the disclosure, a file library containing multiple files can be predetermined, and each file in the file library can be defined as a target file. Steps S110 and S120 are executed sequentially for each target file to determine the target file belonging to the test question file; for the target file belonging to the test question file, step S130 is executed to complete the classification of the target file.

[0055] Figure 2 This illustration shows a flowchart of another test question file screening method provided in an embodiment of this disclosure, such as... Figure 2 As shown, the method mainly includes the following steps:

[0056] S210: Extract feature information from multiple dimensions of the target file to be identified.

[0057] In this disclosure, the dimensions of the target file include title, abstract, body, and document format dimensions, and may also include other dimensions. The type of feature information corresponding to each dimension can be pre-defined. This feature information may, to some extent, reflect whether the target file is a test document. For example, feature information may include keywords, high-frequency words, specified sentence structures, and format types. The process of extracting feature information for each dimension of the target file is further described below:

[0058] In this embodiment of the disclosure, for the title dimension of the target file, title keywords are extracted from the title text of the target file to be identified, and the title keywords are used as feature information of the title dimension.

[0059] In this embodiment of the disclosure, for the summary dimension of the target file, summary keywords are extracted from the summary text of the target file to be identified, and the summary keywords are used as feature information for the title dimension.

[0060] Optionally, for the title text and abstract text of the target file, irrelevant symbols, function words and other interfering fields in the text can be removed first. The text can be segmented by punctuation marks (such as commas and periods) to obtain clauses, the clauses can be tokenized to obtain at least one word, and then the words can be re-aggregated and sorted to extract the corresponding keywords.

[0061] It can be understood that if the target file is a test question file, then the keywords extracted from the title text and abstract text of the target file are very likely to include at least one of the text contents such as "test question", "test paper", "examination" and "test"; if the target file is not a test question file, then the possibility that the keywords extracted from the title text and abstract text of the target file contain the above text contents is relatively small.

[0062] In the embodiment of the present disclosure, for the body dimension of the target file, high-frequency words and / or question sentences of the title are extracted from the body of the target file to be recognized; the high-frequency words and / or question sentences are used as the feature information of the body dimension.

[0063] In the embodiment of the present disclosure, words with a frequency of occurrence exceeding a preset frequency in the body of the target file can be used as high-frequency words. Here, it should be noted that in order to avoid determining words without substantial meaning as high-frequency words, the embodiment of the present disclosure can first remove interfering fields such as function words in the body. For example, words such as "of", "already", and "ah" in the body can be removed in advance. It can be understood that if the target file is a test question file, then the high-frequency words extracted from the body of the target file are very likely to include words related to test questions such as "answer", "calculate", "select", "equal to", and "each question is X points"; if the target file is not a test question file, then the possibility that the high-frequency words extracted from the body of the target file include the above words is relatively small.

[0064] Optionally, in the embodiment of the present disclosure, question sentences of the title can be extracted from the body of the target file according to information such as the text segmentation method, word nature, and common symbols (such as question marks) of questions in the body of the target file. It can be understood that if the target file is a test question file, then the question sentences of the title extracted from the body of the target file can include questions such as "select which one", "which of the following is correct", and "what is the value equal to"; if the target file is not a test question file, then the possibility of extracting question sentences of the title from the body of the target file is relatively small, and at this time, the feature information of the body dimension of the target file may not include question sentences of the title.

[0065] In the embodiment of the present disclosure, for the file format dimension of the target file, the format type of the target file to be recognized is detected; the format type is used as the feature information of the file format dimension.

[0066] File formats typically include Word, PDF, Excel, and PPT. This embodiment of the disclosure can detect which of these formats the target file to be identified is specifically in, and use the detected format type as feature information of the target file's file format dimension. It can be understood that if the target file is not a test question file, then the target file format is usually Word or PDF; when the target file format is Excel or PPT, then the probability that the target file is a test question file is usually relatively low.

[0067] In this embodiment of the disclosure, the dimension of the target file also includes a file type dimension. For the file type dimension of the target file, the file type of the target file to be identified is detected; the file type is used as feature information of the file type dimension.

[0068] To facilitate file management, target files are typically stored in a file library according to file type. For example, file types may include science and technology, lifestyle, entertainment, music, history, and exams. This embodiment of the disclosure can detect which of these types the target file to be identified belongs to, and use the detected file type as feature information of the target file's file type dimension. It can be understood that if the target file's file type is exam-related, then the target file is highly likely to be a test question file; if the target file's file type is other than exam-related, then the probability that the target file is a test question file is relatively low.

[0069] S220: Input multi-dimensional feature information into a pre-trained test item recognition model. For each dimension of feature information, use the test item recognition model to determine the credibility of the target file belonging to the test item file based on the feature information.

[0070] In this embodiment of the disclosure, the test item recognition model can determine a level of credibility for a target file to belong to a test item file based on each feature information. As mentioned above, the dimensions of the target file in this embodiment include five dimensions: title dimension, abstract dimension, body text dimension, file format dimension, and file type dimension. By inputting the feature information of these five dimensions into the pre-trained test item recognition model, the test item recognition model can determine five levels of credibility corresponding to the feature information of the five dimensions.

[0071] It is understandable that the credibility of feature information in different dimensions may vary. Taking the content of feature information in each dimension introduced in step S220 as an example, for the title text and abstract text of the target file, if its feature information contains at least one of the following: "test question", "test paper", "exam", and "test", its credibility is relatively high; for the body text dimension of the target file, when its feature information contains words and / or question sentences related to test questions, its credibility is relatively high; for the file format dimension of the target file, when its feature information contains Word or PDF format information, its credibility is relatively high; for the file type dimension of the target file, when its feature information contains exam-related information, its credibility is relatively high.

[0072] S230: Based on the credibility of the feature information of multiple dimensions, determine whether the target file belongs to the test question file.

[0073] After determining multiple credibility levels corresponding to feature information of multiple dimensions, the embodiments of this disclosure can combine these multiple credibility levels to finally determine whether the target file belongs to the test question file, thereby ensuring the objectivity and reliability of the recognition result.

[0074] Optionally, embodiments of this disclosure can perform a weighted summation of the credibility scores corresponding to feature information from multiple dimensions to obtain a comprehensive credibility score. Based on the comprehensive credibility score and a preset credibility threshold, the identification result of whether the target file belongs to the test question file is determined. Specifically, for each dimension of feature information, the credibility score and weight corresponding to that feature information are multiplied to obtain the corresponding product; the comprehensive credibility score is obtained by summing multiple products. When the comprehensive credibility score is greater than the preset credibility threshold, the identification result of the target file belonging to the test question file is determined; when the comprehensive credibility score is not greater than the preset credibility threshold, the identification result of the target file not belonging to the test question file is determined.

[0075] It should be noted that the credibility threshold can be determined according to actual design needs; for example, the credibility threshold can be set to 0.9. The credibility weight corresponding to the feature information of each dimension is obtained by training the test question recognition model. Generally speaking, the credibility weight corresponding to the feature information of the title dimension is the largest, while the credibility weight corresponding to the feature information of other dimensions is smaller. This embodiment sets different weights for the credibility corresponding to the feature information of different dimensions, intuitively reflecting the importance of each dimension of feature information to the recognition result, and ensuring the objectivity of the recognition result.

[0076] S240: When the target file is a test question file, extract at least one of the following from the target file: regional information, school level information, and subject information, as test question tags.

[0077] In this embodiment of the disclosure, information such as schools and administrative regions can be extracted from the target file as tags for regional information; grades (such as first grade of primary school, first grade of junior high school, first year of primary school and first year of university) can be extracted from the target file as tags for educational stage information; and test subjects from the target file, such as (mathematics, Chinese, history or biology), can be used as tags for subject information.

[0078] S250: Classify the target document based on at least one of the following: regional information, grade level information, and subject information.

[0079] This embodiment of the disclosure can pre-establish a classification knowledge graph, which contains the relationship between each grade and the corresponding academic stage. The division criteria for academic stages can be determined according to actual design needs. For example, it can be divided into K12 academic stage (primary to high school) and university academic stage. The academic stage type to which the test paper file belongs is determined based on the regional information tags extracted in step S240.

[0080] The classification knowledge graph of this embodiment can also include subordinate relationships between schools and administrative regions, subordinate relationships between administrative regions at different levels, etc., and the regional type (such as district-level region, city-level region, provincial-level region, and national-level region) to which the test paper file belongs is determined by the tags of the regional information class extracted in step S240. Of course, this embodiment can also determine the subject type to which the test paper file belongs based on subject information. It can be understood that after classifying the target files, the target files can be stored according to categories, thereby facilitating the summarization and management of the selected target files by type.

[0081] In this embodiment of the disclosure, after the test question files are selected, the test question files can be manually reviewed to determine whether the test question files selected by the test question recognition model are real test question files. Based on the review results, the parameters of the test question recognition model are adjusted to improve the accuracy of the recognition results of the test question recognition model.

[0082] In this embodiment of the disclosure, after classifying the test question files, the classification results of the test question files can be reviewed, and the classification knowledge graph can be corrected based on the review results, thereby improving the accuracy of the classification results.

[0083] Figure 3 This illustration shows a flowchart of a test item recognition model training method provided in an embodiment of the present disclosure, as shown below. Figure 3 As shown, the method mainly includes the following steps:

[0084] S310: Indicate whether each sample file in multiple sample files belongs to the actual result of the test question file.

[0085] It should be noted that multiple sample files can include test question files and non-test question files. After labeling the actual results of whether each sample file in the multiple sample files belongs to the test question file, these sample files can be used as the training set of the test question recognition model.

[0086] S320: Extract feature information from multiple dimensions of each sample file from multiple sample files.

[0087] In this implementation, the dimensions of the sample file include title, abstract, body, and file format dimensions, and may also include other dimensions. The type of feature information corresponding to each dimension can be pre-defined. This feature information may, to some extent, reflect whether the sample file is a test question file. For example, feature information may include keywords, high-frequency words, specified sentence structures, and format types. Of course, the dimensions of the sample file may also include other dimensions, such as file type. Here, the specific steps for extracting the feature information of multiple dimensions of the sample file in step S320 can refer to the steps for extracting the feature information of multiple dimensions of the target file in step S210 above, and will not be repeated here.

[0088] S330: Input multi-dimensional feature information into the test item recognition model, and use the test item recognition model to output the recognition result of whether the sample file belongs to the test item file.

[0089] In this embodiment of the disclosure, the test item recognition model can determine a level of confidence that a sample file belongs to a test item file based on each feature information. As mentioned above, the dimensions of the sample file in this embodiment include five dimensions: title dimension, abstract dimension, body text dimension, file format dimension, and file type dimension. By inputting the feature information of these five dimensions into the pre-trained test item recognition model, the test item recognition model can determine five levels of confidence corresponding to the feature information of the five dimensions.

[0090] Optionally, in this embodiment of the disclosure, for each dimension of feature information, the test question recognition model is used to determine the credibility of the sample file belonging to the test question file based on the feature information; based on the credibility corresponding to the feature information of multiple dimensions, the recognition result of whether the sample file belongs to the test question file is determined.

[0091] Optionally, embodiments of this disclosure can perform a weighted summation of the credibility levels corresponding to feature information in multiple dimensions to obtain a comprehensive credibility level. Specifically, for feature information in each dimension, the credibility level corresponding to that feature information is multiplied by its weight to obtain a corresponding product; the comprehensive credibility level is obtained by summing multiple products.

[0092] After obtaining the overall credibility score, the identification result of whether the sample file belongs to the test question file can be determined based on the overall credibility score and a preset credibility threshold. Specifically, when the overall credibility score is greater than the preset credibility threshold, the identification result of the sample file belonging to the test question file is determined; when the overall credibility score is not greater than the preset credibility threshold, the identification result of the sample file not belonging to the test question file is determined.

[0093] S340: Based on the actual results and recognition results of each sample file, determine the recognition accuracy of the test item recognition model.

[0094] This disclosure determines whether the actual result and the recognition result of the sample file are consistent. If they are consistent, the recognition result is correct; if they are inconsistent, the recognition result is incorrect. The accuracy rate of the test item recognition model can be obtained by statistically analyzing the accuracy rates of the recognition results of multiple sample files.

[0095] S350: Adjust the parameters of the test item recognition model based on the recognition accuracy and the preset accuracy threshold.

[0096] In this embodiment of the disclosure, the accuracy threshold can be determined according to actual design needs. For example, the accuracy threshold can be set to 98%. It is understood that the purpose of adjusting the parameters of the test item recognition model is to make the recognition accuracy of the test item recognition model greater than or equal to the accuracy threshold. When the recognition accuracy of the test item recognition model is greater than or equal to the accuracy threshold, the training process can be terminated.

[0097] Optionally, in adjusting the parameters of the test item recognition model, embodiments of this disclosure may adjust the weight of the credibility corresponding to the feature information of each dimension in the test item recognition model.

[0098] Based on the same principle as the test document screening method mentioned above, Figure 4 A schematic diagram of a test question document screening device provided in an embodiment of this disclosure is shown. Figure 4 As shown, the test question file screening device 400 includes a target feature extraction module 410, a target result output module 420, and a test question classification module 430.

[0099] The target feature extraction module 410 is used to extract feature information of multiple dimensions of the target file to be identified. The dimensions of the target file include title dimension, abstract dimension, body text dimension and file format dimension.

[0100] The target result output module 420 is used to determine whether the target file belongs to the test question file based on feature information of multiple dimensions.

[0101] The test question classification module 430 is used to extract at least one preset test question tag information from the target file when the target file is a test question file, and classify the target file based on the test question tag information.

[0102] The test question file screening device provided in this embodiment comprehensively identifies whether a file belongs to test question files by integrating information from multiple dimensions such as the file title, abstract, body text, and file format. For target files that belong to test question files, classification can be automatically completed based on their contained tag information. The above process can significantly improve the accuracy of file recognition results, largely avoiding the omission of real test question files due to misjudgment. The test question file screening process can be completed automatically, which not only improves the efficiency of test question screening but also reduces labor costs.

[0103] In this embodiment of the disclosure, when the target result output module 420 is used to determine whether a target file belongs to a test question file based on feature information of multiple dimensions, it is specifically used to: input feature information of multiple dimensions into a pre-trained test question recognition model, and use the test question recognition model to output the recognition result of whether the target file belongs to a test question file.

[0104] In this embodiment of the disclosure, when the target result output module 420 is used to output the identification result of whether the target file belongs to the test question file using the test question identification model, it is specifically used for:

[0105] For each dimension of feature information, the test item recognition model is used to determine the credibility of the target file belonging to the test item file based on the feature information;

[0106] Based on the credibility of the feature information from multiple dimensions, the identification result of whether the target file belongs to the test question file is determined.

[0107] In this embodiment of the disclosure, when the target result output module 420 determines whether a target file belongs to a test question file based on the credibility corresponding to the feature information of multiple dimensions, it is specifically used for:

[0108] The overall credibility is obtained by weighted summing of the credibility values ​​corresponding to the feature information of multiple dimensions;

[0109] Based on the overall credibility and the preset credibility threshold, the identification result of whether the target file belongs to the test question file is determined;

[0110] The credibility weights corresponding to the feature information of each dimension are obtained by training the test item recognition model.

[0111] In this embodiment of the disclosure, the target feature extraction module 410, when extracting feature information of multiple dimensions of the target file to be identified, is specifically used for:

[0112] For the title dimension of the target file, title keywords are extracted from the title text of the target file to be identified, and the title keywords are used as feature information of the title dimension.

[0113] And / or, for the summary dimension of the target file, extract summary keywords from the summary text of the target file to be identified, and use the summary keywords as feature information for the title dimension.

[0114] In this embodiment of the disclosure, the target feature extraction module 410, when extracting feature information of multiple dimensions of the target file to be identified, is specifically used for:

[0115] For the text dimension of the target document, high-frequency words and / or questions are extracted from the text of the target document to be identified;

[0116] High-frequency words and / or questions are used as feature information in the main text dimension.

[0117] In this embodiment of the disclosure, the target feature extraction module 410, when extracting feature information of multiple dimensions of the target file to be identified, is specifically used for:

[0118] Based on the file format dimension of the target file, the format type of the target file to be identified is detected;

[0119] The format type is used as a feature information of the file format dimension.

[0120] In this embodiment of the disclosure, the dimensions of the target file also include the file type dimension; when the target feature extraction module 410 is used to extract feature information of multiple dimensions of the target file to be identified, it is specifically used for:

[0121] Based on the file type dimension of the target file, the file type of the target file to be identified is detected;

[0122] File type is used as a feature information of the file type dimension.

[0123] In this embodiment of the disclosure, when the test question classification module 430 is used to extract at least one preset test question tag information from the target file, it is specifically used to: extract at least one of the regional information, school level information and subject information from the target file as test question tags.

[0124] It is understood that the modules of the test question file screening device in this embodiment have the function of implementing the corresponding steps of the test question file screening method described above. This function can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. These modules can be software and / or hardware; each module can be implemented individually or multiple modules can be integrated. For a detailed description of the functions of each module of the test question file screening device, please refer to the corresponding description of the test question file screening method described above, which will not be repeated here.

[0125] Based on the same principle as the test item recognition model training method mentioned above, Figure 5 A schematic diagram of a test item recognition model training device provided in an embodiment of this disclosure is shown. Figure 5 As shown, the test item recognition model training device 500 includes a result labeling module 510, a sample feature extraction module 520, a sample result output module 530, an accuracy determination module 540, and a parameter adjustment module 550.

[0126] The result annotation module 510 is used to annotate whether each sample file in multiple sample files is the actual result of the test question file.

[0127] The sample feature extraction module 520 is used to extract feature information of each sample file in multiple dimensions. The dimensions of the sample files include title dimension, abstract dimension, body text dimension and file format dimension.

[0128] The sample result output module 530 is used to input multi-dimensional feature information into the test item recognition model and use the test item recognition model to output the recognition result of whether the sample file belongs to the test item file.

[0129] The accuracy determination module 540 is used to determine the recognition accuracy of the test item recognition model based on the actual results and recognition results of each sample file.

[0130] The parameter adjustment module 550 is used to adjust the parameters of the test item recognition model based on the recognition accuracy and a preset accuracy threshold.

[0131] In this embodiment of the disclosure, when the sample result output module 530 is used to output the recognition result of whether a sample file belongs to a test question file using the test question recognition model, it is specifically used for:

[0132] For each dimension of feature information, the test item recognition model is used to determine the credibility of a sample file belonging to a test item file based on the feature information;

[0133] Based on the credibility of the feature information from multiple dimensions, the identification result of whether the sample file belongs to the test question file is determined.

[0134] In this embodiment of the disclosure, when the sample result output module 530 determines whether a sample file belongs to a test question file based on the credibility corresponding to the feature information of multiple dimensions, it is specifically used to: perform a weighted summation of the credibility corresponding to the feature information of multiple dimensions to obtain a comprehensive credibility; and determine whether a sample file belongs to a test question file based on the comprehensive credibility and a preset credibility threshold.

[0135] When adjusting the parameters of the test item recognition model, the parameter adjustment module 550 is specifically used to adjust the confidence weight of the feature information corresponding to each dimension in the test item recognition model.

[0136] It is understood that the modules of the test item recognition model training device in this embodiment have the function of implementing the corresponding steps of the test item recognition model training method described above. This function can be implemented by hardware or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions. These modules can be software and / or hardware; each module can be implemented individually or multiple modules can be integrated. For a detailed description of the functions of each module of the test item recognition model training device, please refer to the corresponding description of the test item recognition model training method described above, which will not be repeated here.

[0137] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0138] Figure 6 A schematic block diagram of an example electronic device is shown that can be used to implement the test item file screening method or test item recognition model training method provided in embodiments of this disclosure. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0139] like Figure 6As shown, the electronic device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. The RAM 603 may also store various programs and data required for the operation of the electronic device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0140] Multiple components in electronic device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of displays, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows electronic device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0141] The computing unit 601 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the test item file screening method or the test item recognition model training method. For example, in some embodiments, the test item file screening method or the test item recognition model training method can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the test item file screening method or the test item recognition model training method described above can be performed. Alternatively, in other embodiments, the computing unit 601 may be configured by any other suitable means (e.g., by means of firmware) to perform a test item file screening method or a test item recognition model training method.

[0142] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0143] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0144] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0145] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0146] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0147] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0148] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0149] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A test question file screening method, comprising: extracting feature information of multiple dimensions of a target file to be identified, the dimensions of the target file including a title dimension, an abstract dimension, a body dimension, a file type dimension, and a file format dimension; wherein the feature information of the title dimension at least includes title keywords; the feature information of the abstract dimension at least includes abstract keywords; the feature information of the body dimension includes high-frequency words and / or a question sentence; the feature information of the file type dimension at least includes a file type; and the feature information of the file format dimension at least includes a format type; determining whether the target file belongs to a test question file based on the feature information of the multiple dimensions; when the target file is a test question file, extracting at least one preset test question label information from the target file, and classifying the target file based on the test question label information; wherein the extracting feature information of multiple dimensions of a target file to be identified comprises: for the body dimension of the target file, extracting a question sentence from the body of the target file to be identified; wherein the question sentence is extracted from the body of the target file according to a text segmentation manner, a word part of speech, and common symbols of a question in the body of the target file; taking the question sentence as the feature information of the body dimension; for the file format dimension of the target file, detecting a format type of the target file to be identified; wherein the format type includes at least one of a word processing document, a portable document format, a spreadsheet document, and a presentation document; taking the format type as the feature information of the file format dimension; for the file type dimension of the target file, detecting a file type of the target file to be identified; wherein the file type includes at least one of a science and technology category, a life category, an entertainment category, a music category, a history category, and an examination category; taking the file type as the feature information of the file type dimension; wherein the extracting at least one preset test question label information from the target file comprises: extracting at least one of region information, grade information, and subject information from the target file as a test question label; the classifying the target file based on the test question label information comprises: classifying the target file based on at least one of the region information, the grade information, and the subject information; the region information includes at least one of a school and an administrative region; the grade information includes a grade; and the subject information includes a test subject; the method further comprises: pre-establishing a classification knowledge graph, the classification knowledge graph including a relationship between at least one of the region information, the grade information, and the subject information and a classification type.

2. The method of claim 1, wherein, the determining whether the target file belongs to a test question file based on the feature information of the multiple dimensions comprises: inputting the feature information of the multiple dimensions into a pre-trained test question identification model, and outputting, by the test question identification model, a determination result of whether the target file belongs to a test question file.

3. The method of claim 2, wherein, The identification result of whether the target file belongs to a test question file is output by the test question identification model, and includes: For each dimension of feature information, the test question identification model determines a credibility that the target file belongs to a test question file according to the feature information; According to the credibility corresponding to the feature information of each dimension, the identification result of whether the target file belongs to a test question file is determined.

4. The method of claim 3, wherein, The identification result of whether the target file belongs to a test question file is determined according to the credibility corresponding to the feature information of each dimension, and includes: The credibility corresponding to the feature information of each dimension is weighted and summed to obtain a comprehensive credibility; Based on the comprehensive credibility and a preset credibility threshold, the identification result of whether the target file belongs to a test question file is determined; The weight of the credibility corresponding to the feature information of each dimension is obtained by training the test question identification model.

5. The method of any one of claims 1 to 4, wherein, The feature information of multiple dimensions of the target file to be identified is extracted, and includes: For the title dimension of the target file, a title keyword is extracted from the title text of the target file to be identified, and the title keyword is used as the feature information of the title dimension; And / or, for the abstract dimension of the target file, an abstract keyword is extracted from the abstract text of the target file to be identified, and the abstract keyword is used as the feature information of the title dimension.

6. The method of any one of claims 1 to 4, wherein, The feature information of multiple dimensions of the target file to be identified is extracted, and includes: For the body dimension of the target file, a high-frequency word is extracted from the body of the target file to be identified; The high-frequency word is used as the feature information of the body dimension.

7. A test question identification model training method, comprising: annotating the actual result of whether each sample file in a plurality of sample files belongs to a test question file; extracting feature information of multiple dimensions of each sample file in the plurality of sample files, the dimensions of the sample file including a title dimension, an abstract dimension, a body dimension, a file type dimension, and a file format dimension; wherein the feature information of the title dimension includes at least a title keyword; the feature information of the abstract dimension includes at least an abstract keyword; the feature information of the body dimension includes a high-frequency word and / or a question sentence; the feature information of the file type dimension includes at least a file type; and the feature information of the file format dimension includes at least a format type; inputting the feature information of the multiple dimensions into a test question identification model, and outputting an identification result of whether the sample file belongs to a test question file by the test question identification model; determining an identification accuracy of the test question identification model based on the actual result and the identification result of each sample file; adjusting parameters of the test question identification model based on the identification accuracy and a preset accuracy threshold.

8. The method of claim 7, wherein, The identification result of whether the sample file belongs to a test question file is output by the test question identification model, and includes: For each dimension of feature information, the test question identification model determines a credibility that the target file belongs to a test question file according to the feature information; According to the credibility corresponding to the feature information of multiple dimensions respectively, a recognition result of whether the sample file belongs to a test question file is determined.

9. The method of claim 8, wherein, According to the credibility corresponding to the feature information of multiple dimensions respectively, a recognition result of whether the sample file belongs to a test question file is determined, including: weighting and summing the credibility corresponding to the feature information of multiple dimensions respectively to obtain a comprehensive credibility; based on the comprehensive credibility and a preset credibility threshold, a recognition result of whether the sample file belongs to a test question file is determined; The parameter of the test question recognition model is adjusted, including: adjusting the weight of the credibility corresponding to the feature information of each dimension in the test question recognition model.

10. A test question file screening device, comprising: A target feature extraction module is configured to extract feature information of multiple dimensions of a target file to be recognized, wherein the dimensions of the target file include a title dimension, an abstract dimension, a text dimension, a file type dimension, and a file format dimension; wherein the feature information of the title dimension includes at least a title keyword; the feature information of the abstract dimension includes at least an abstract keyword; the feature information of the text dimension includes a high-frequency word and / or a question sentence; the feature information of the file type dimension includes at least a file type; and the feature information of the file format dimension includes at least a format type; A target result output module is configured to determine a recognition result of whether the target file belongs to a test question file based on the feature information of the multiple dimensions; A test question classification module is configured to extract at least one preset test question label information from the target file when the target file is a test question file, and classify the target file based on the test question label information; The target feature extraction module is specifically configured to extract a question sentence from the text of the target file in the text dimension of the target file to be recognized; wherein the question sentence is extracted from the text of the target file according to a text segmentation manner, a word part of speech, and a common symbol of a question in the text of the target file; The question sentence is used as the feature information of the text dimension; For the file format dimension of the target file, a format type of the target file to be recognized is detected; wherein the format type includes at least one of a word processing document, a portable document format, a spreadsheet document, and a presentation document; The format type is used as the feature information of the file format dimension; For the file type dimension of the target file, a file type of the target file to be recognized is detected; wherein the file type includes at least one of a science and technology type, a life type, an entertainment type, a music type, a history type, and an examination type; The file type is used as the feature information of the file type dimension; The test question classification module is specifically configured to: Extract at least one of regional information, grade information, and subject information from the target file as a test question label. classify the target file based on at least one of the region information, the school stage information, and the subject information; the region information comprises at least one of a school and an administrative region; the school stage information comprises a grade; and the subject information comprises a subject of a test question; the apparatus further comprises a graph module configured to pre-establish a classification knowledge graph, the classification knowledge graph comprising a relationship between at least one of the region information, the school stage information, and the subject information and a classification type.

11. The apparatus of claim 10, wherein, The target feature extraction module is configured to: input the feature information of the multiple dimensions into a pre-trained test question recognition model, and output, by using the test question recognition model, the recognition result of whether the target file belongs to a test question file.

12. The apparatus of claim 11, wherein, The target result output module is configured to: determine, by using the test question recognition model, a credibility of the target file belonging to a test question file according to the feature information of each dimension; determine the recognition result of whether the target file belongs to a test question file according to the credibilities corresponding to the feature information of the multiple dimensions respectively.

13. The apparatus of claim 12, wherein, The target result output module is configured to: weight and sum the credibilities corresponding to the feature information of the multiple dimensions respectively to obtain a comprehensive credibility; determine the recognition result of whether the target file belongs to a test question file based on the comprehensive credibility and a preset credibility threshold; wherein the weight of the credibility corresponding to the feature information of each dimension is obtained by training the test question recognition model.

14. A test question recognition model training apparatus, comprising: a result labeling module configured to label an actual result of whether each sample file in a plurality of sample files belongs to a test question file; a sample feature extraction module configured to extract feature information of multiple dimensions of each sample file in the plurality of sample files, the dimensions of the sample file comprising a title dimension, an abstract dimension, a body dimension, a file type dimension, and a file format dimension; wherein the feature information of the title dimension comprises at least a title keyword; the feature information of the abstract dimension comprises at least an abstract keyword; the feature information of the body dimension comprises a high-frequency word and / or a question sentence; the feature information of the file type dimension comprises at least a file type; and the feature information of the file format dimension comprises at least a format type; a sample result output module configured to input the feature information of the multiple dimensions into a test question recognition model, and output, by using the test question recognition model, a recognition result of whether the sample file belongs to a test question file; an accuracy rate determination module configured to determine a recognition accuracy rate of the test question recognition model based on the actual result and the recognition result of each sample file. The parameter adjustment module is configured to adjust parameters of the test question recognition model based on the recognition accuracy and a preset accuracy threshold. 15.An electronic device, comprising: at least one processor; and a memory connected to the at least one processor in communication; 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 method of any one of claims 1-6 or claims 7-9.

16. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method of any one of claims 1-6 or claims 7-9. 17.A computer program product comprising a computer program which, when executed by a processor, implements the method of any one of claims 1-6 or claims 7-9.

Images