File processing method and device, file identification method and device, equipment and storage medium

The method improves AI-generated file recognition by extracting and optimizing template features using genetic algorithms and AB experiments, addressing mismatches and errors in AI documents.

CN119089860AActive Publication Date: 2024-12-06BAIDU ONLINE NETWORK TECH (BEIJIBG) CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411067237.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-12-06
Estimated Expiration
2044-08-05

AI Technical Summary

Technical Problem

Existing AI-generated documents face issues such as mismatches between images and texts, factual errors, and watered-down content, necessitating improved methods for identifying and distinguishing them from non-AI-generated documents.

Method used

A method involving extracting first features from a file template, combining them to form second features, and using genetic algorithms or AB experiments to optimize these features for accurate recognition, along with a matching process to determine AI-generated files.

Benefits of technology

Enhances the accuracy and efficiency of identifying AI-generated files by constructing strong features that account for template variations, reducing misidentification and improving recognition speed and quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089860A_ABST
    Figure CN119089860A_ABST
Patent Text Reader

Abstract

The invention provides a file processing method and device, a file identification method and device, equipment and a storage medium, and relates to the technical field of data processing, in particular to the fields of artificial intelligence, file identification and the like. The specific implementation scheme is as follows: extracting a plurality of first features from a file template required for generating a target type file based on an artificial intelligence technology; and combining the plurality of first features to screen at least one second feature capable of identifying a file page generated based on an artificial intelligence technology. According to the embodiment of the invention, the template features and the combined feature rules used for identifying the artificial intelligence technology generated file are screened out, so that the identification accuracy and the identification efficiency of the artificial intelligence technology generated file can be improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to the fields of artificial intelligence, file recognition, etc. Background Art

[0002] Artificial Intelligence (AI) technology is a branch of computer science. AI attempts to understand the essence of things and is able to make logical inferences. With the development of AI technology, the implementation of AI technology in different fields is particularly important.

[0003] At present, AI technology can understand knowledge, sort out data, and automatically generate files based on its powerful logical reasoning ability. For example, it can write PPT (PowerPoint, presentation) documents to improve user work efficiency. Summary of the invention

[0004] The present disclosure provides a file processing method, a file identification method, an apparatus, a device and a storage medium.

[0005] According to one aspect of the present disclosure, there is provided a file processing method, comprising:

[0006] Extracting a plurality of first features from a file template required for generating a target type file based on artificial intelligence technology;

[0007] The plurality of first features are combined to filter at least one second feature capable of identifying a document page generated based on artificial intelligence technology.

[0008] According to one aspect of the present disclosure, there is provided a file identification method, comprising:

[0009] Extracting a plurality of third features from the to-be-identified pages of the to-be-identified file;

[0010] Combining multiple third features to obtain at least one fourth feature;

[0011] Matching the plurality of third features and at least one fourth feature with a known feature set of a file template required for generating a file type of a file to be identified based on artificial intelligence technology to obtain a feature matching degree;

[0012] When the feature matching degree meets the preset conditions, it is determined that the page to be identified is generated based on artificial intelligence technology.

[0013] According to another aspect of the present disclosure, there is provided a file processing device, comprising:

[0014] A first extraction module, used to extract a plurality of first features from a file template required for generating a target type file based on artificial intelligence technology;

[0015] The screening module is used to combine multiple first features to screen at least one second feature that can identify a file page generated based on artificial intelligence technology.

[0016] According to another aspect of the present disclosure, there is provided a file identification device, comprising:

[0017] A second extraction module is used to extract a plurality of third features from the to-be-identified page of the to-be-identified file;

[0018] A combining module, used for combining a plurality of third features to obtain at least one fourth feature;

[0019] A matching module, used for matching the plurality of third features and at least one fourth feature with a known feature set of a file template required for generating a file type of a file to be identified based on artificial intelligence technology, to obtain a feature matching degree;

[0020] The determination module is used to determine that the page to be identified is generated based on artificial intelligence technology when the feature matching degree meets the preset conditions.

[0021] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0022] at least one processor; and

[0023] a memory communicatively connected to the at least one processor; wherein,

[0024] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute any method in the embodiments of the present disclosure.

[0025] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any method according to the embodiments of the present disclosure.

[0026] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which implements any method according to the embodiments of the present disclosure when executed by a processor.

[0027] The disclosed embodiments can improve the recognition accuracy and efficiency of files generated by artificial intelligence technology by screening out template features and combination feature rules for identifying files generated by artificial intelligence technology.

[0028] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0030] Figure 1 is a flowchart of a file processing method according to an embodiment of the present disclosure;

[0031] Figure 2 is a flow chart of the second feature of the screening sequence according to an embodiment of the present disclosure;

[0032] Figure 3 is a flowchart of a file identification method according to an embodiment of the present disclosure;

[0033] Figure 4 This is a schematic diagram of the process of identifying a PPT file generated by artificial intelligence technology according to an embodiment of the present disclosure.

[0034] Figure 5 is a structural schematic diagram of a file processing device according to an embodiment of the present disclosure;

[0035] Figure 6 is a structural schematic diagram of a file identification device according to an embodiment of the present disclosure;

[0036] Figure 7 It is a block diagram of an electronic device used to implement the file processing method and / or file identification method of the embodiments of the present disclosure. DETAILED DESCRIPTION

[0037] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0038] The terms "first", "second", etc. in this disclosure are used to distinguish similar objects, and are not necessarily used to describe a specific order or precedence. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, including a series of steps or units. The method, system, product or device is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0039] With the application of artificial intelligence technology in document generation, the demand for the generation of personalized documents is growing. However, documents generated by artificial intelligence technology still have potential problems such as mismatch between images and texts, factual errors, and watered-down content. Therefore, certain technologies are needed to identify AI-generated files. For example, by identifying AI-generated files, it is easier to provide higher-quality file resources for later upper-level businesses such as library recommendations and searches.

[0040] With the rapid development of network informatization, the ability of AI file recognition should be fast and efficient. In view of this, the embodiment of the present disclosure provides a file processing method. In this method, based on the fact that the generation of files by artificial intelligence technology needs to rely on file templates, it is proposed to identify the file content generated by artificial intelligence technology by extracting the template features of the file template.

[0041] like Figure 1 FIG. 1 is a flowchart of a file processing method proposed in an embodiment of the present disclosure, including:

[0042] S101, extracting a plurality of first features from a file template required for generating a target type file based on artificial intelligence technology.

[0043] A file template refers to a pre-designed document structure, which contains various pre-defined template elements, such as text boxes, pictures, tables, graphics for expressing hierarchical structures, elements for beautification, etc. Target type files generated by artificial intelligence technology often retain template features, such as some layout structures, fonts, specific elements of theme styles, etc. These template features are extracted as the first features. It can be seen that the first features are some weak features, and multiple first features can be extracted based on the file template to describe the features of the file template from different dimensions and angles.

[0044] However, only relying on these weak features to match the file to be identified to identify the file generated based on artificial intelligence technology may result in a failure to match or a mismatch. In addition, if some elements in the file template change, such as displacement and deformation, matching based on these weak features alone will result in a mismatch. Therefore, in the embodiment of the present disclosure, in S102, multiple first features are combined to screen at least one second feature that can identify the file page generated based on artificial intelligence technology.

[0045] The extracted first features are combined in a specific combination manner, and combined features that can effectively identify file pages generated based on artificial intelligence are screened out as second features.

[0046] These second features can be understood as strong features. As the name implies, strong features are features obtained by combining weak features, and can further express the characteristics of the file template compared to weak features.

[0047] In summary, in the disclosed embodiments, multiple first features are extracted from the file template required to generate the target type file based on artificial intelligence technology, which can reflect the characteristics of the file generated by artificial intelligence technology from multiple dimensions and angles, thereby ensuring the basic feature requirements for recognition. Furthermore, by combining the first features, more complex and sophisticated recognition features can be constructed, thereby constructing strong features to reduce the possibility of misidentification. At least one second feature that can identify the file page generated based on artificial intelligence technology is screened out, and more qualified combined features can be screened out from a large number of features, and can be compatible with the deformation, displacement changes and other characteristics of some elements in the file template, thereby improving the recognition efficiency of the file page generated based on artificial intelligence technology.

[0048] In summary, the first feature in the embodiment of the present disclosure can be summarized as being used to describe at least one of the typeset features of each of multiple template elements in a file template, the relative features between different template elements, and the relative features between the same type of template elements.

[0049] The typesetting characteristics of multiple template elements in a file template refer to the layout and arrangement of the template elements in the file template, including but not limited to the position, size, direction, alignment, etc. of the template elements. For example, in a PPT template, the starting and ending coordinates of a text box are [(x1, y1), (x2, y2)], the title element may be located at the top center of the page, and the body paragraph may be left-aligned.

[0050] Relative characteristics between different template elements usually refer to the characteristics or attributes obtained by comparing with a reference point or reference value. In different types of files and situations, relative characteristics may have different meanings and applications. For example, in a PPT template, the number of graphics, the size ratio between different graphic types, the concentric circle relationship, transparency, alignment relationship, relative position relationship, etc. can all be used as relative characteristics.

[0051] The relative characteristics between similar template elements refer to the fact that a document template contains multiple similar elements, such as multiple images or multiple titles. Their relative characteristics may include the spacing, alignment, and whether there is a consistent style between them. For another example, a document template includes multiple circles, and the relative relationship between these multiple circles.

[0052] In the disclosed embodiment, the first feature can describe the element features of the file template from multiple dimensions, which helps to improve the accuracy of subsequent identification of whether the file template is generated by artificial intelligence technology. Through these element features of the file template, it is possible to better distinguish between files generated by artificial intelligence technology and files generated by non-artificial intelligence technology, reducing the possibility of misjudgment.

[0053] In the disclosed embodiment, the first features may be randomly combined to obtain at least one second feature.

[0054] In order to improve the quality of file recognition, in the embodiment of the present disclosure, multiple first features may be combined, and at least one second feature may be screened out based on the recall rate and / or accuracy rate of the file recognition result.

[0055] Among them, the file recognition result refers to the recognition result of whether the page to be recognized is a file generated by artificial intelligence technology.

[0056] The recall rate of file recognition results can be understood as the proportion of all files generated by artificial intelligence technology that are correctly recognized.

[0057] The accuracy of file recognition results can be understood as the proportion of all files to be identified that are correctly distinguished as whether they are generated by artificial intelligence technology.

[0058] Therefore, by screening the second features based on the recall rate and / or the accuracy rate, the second features that are effective for identifying AI-generated files can be screened out. A higher recall rate can ensure that as many second features of files that may be generated by AI technology are discovered as possible, reducing missed judgments; while a higher accuracy rate can ensure that the second features of files that are judged to be generated by AI technology have a higher degree of credibility, reducing the probability of misjudgment.

[0059] In some embodiments, based on the recall rate and / or accuracy of the file recognition result, multiple first features can be combined based on a genetic algorithm to obtain at least one second feature with maximizing the objective function as the criterion; the objective function includes the recall rate and / or accuracy of the file recognition result.

[0060] Among them, the genetic algorithm is a heuristic search method that finds solutions to optimization problems by simulating the evolutionary process in nature. In the embodiment of the present disclosure, the optimal second feature is continuously iterated by using the genetic algorithm combined with the recall rate and / or accuracy rate in the objective function to identify files generated by artificial intelligence technology.

[0061] Among them, when the recall rate and the accuracy rate are used at the same time, the two can be combined into an objective function, which is usually the weighted sum of the two. The setting of the weight can be determined by AB experiments according to the needs of the specific problem. That is, the effect of different second features in identifying artificial intelligence technology to generate files is tested through AB experiments, and the second feature with the best performance is selected. In addition, in order to ensure that the number of files generated by the identified artificial intelligence technology does not fall below a certain acceptable threshold, it is necessary to ensure that the recall rate and the accuracy rate meet certain preset values. For example, in the objective function, it is required that the recall rate needs to be greater than the preset value P, and the accuracy rate needs to be greater than the preset value Q.

[0062] In the disclosed embodiment, the optimization based on the objective function by using a genetic algorithm can comprehensively consider important indicators such as recall rate and / or accuracy rate, thereby improving the accuracy and comprehensiveness of recognition. Based on the feature combination of the genetic algorithm, the possibility of different feature combinations can be more effectively explored to find the second feature by optimizing to improve the recognition effect of whether the file is generated by artificial intelligence technology.

[0063] The specific implementation of combining multiple first features based on the genetic algorithm to obtain at least one second feature can be as follows: Figure 2 As shown, including the following:

[0064] S201, randomly combine multiple first features to obtain multiple candidate combination features.

[0065] Select some or all of the available first features for random combination to obtain multiple candidate combination features. For example, "combination feature Z1 = feature 1 & feature 7", that is, feature 1 and feature 7 in the first feature are randomly combined to form candidate combination feature Z1. "Combination feature Z2 = feature 2 & feature 3 & feature 5", that is, feature 2, feature 3 and feature 5 in the first feature are randomly combined to form candidate combination feature Z2.

[0066] S202: Determine the function value of the objective function based on each candidate combination feature.

[0067] For each candidate combination feature generated in S201, its function value is obtained by calculating the objective function. The objective function is usually a standard used to measure the quality of the solution in the optimization problem. In different application scenarios, the objective function can take different forms, such as indicators such as accuracy and / or recall rate. By calculating the function value of each candidate combination feature, the effect of these candidate combination features in identifying whether the file is generated by artificial intelligence technology can be evaluated.

[0068] S203, selecting a plurality of seed combination features from a plurality of candidate combination features according to the size of the function value.

[0069] The function values ​​of the candidate combination features are sorted from large to small, and the top M best candidate combination features are selected as seed combination features, where M is a positive integer.

[0070] S204, performing crossover and mutation operations on the multiple seed combination features to obtain multiple new candidate combination features, and returning to execute the step of respectively determining the function value of the objective function based on each candidate combination feature, until at least one second feature that maximizes the objective function is screened out.

[0071] For the selected seed combination features, crossover and mutation operations are performed, where:

[0072] The crossover operation refers to exchanging some features of two or more seed combination features to generate a new combination feature. For example, feature 7 in combination feature 1 and feature 5 in combination feature 2 are exchanged to obtain a new candidate combination feature 1 = feature 1 & feature 5. The new candidate combination feature Z2 = feature 2 & feature 3 & feature 7.

[0073] The mutation operation is to make small changes to certain features in the seed combination features, such as adding, deleting or replacing a certain feature value, where the range of mutation should follow the range of mutation in practical applications. For example, the range of variation of different attributes of the same element in a large number of files generated by artificial intelligence technology can be counted. Then, based on the range of variation, the corresponding attribute values ​​are randomly perturbed. For example, the displacement, deformation and other features of the same template element are counted to obtain a general displacement variation range and deformation variation range. Then, feature values ​​are randomly extracted from the range with a higher probability of appearing in the displacement variation range, and the displacement attributes in the original feature are updated to achieve mutation. The processing of other attributes is similar and will not be repeated here.

[0074] In summary, through crossover and mutation operations, new candidate combination features can be generated to make the objective function converge as much as possible.

[0075] After obtaining new candidate combination features, S202 and S203 can be repeated, that is, the function value of the objective function is calculated again for the newly generated candidate combination features, and a new batch of seed combination features are screened out according to the function value. This process will be repeated, and the combination method will be continuously optimized until at least one second feature that maximizes the objective function is found.

[0076] In the disclosed embodiment, a large number of candidate combination features can be generated by randomly combining multiple first features, which can ensure the comprehensiveness of the combination features used to identify whether a file is generated by artificial intelligence technology. By calculating the function value of the target function for each candidate combination feature and screening according to the size of the function value, invalid or inefficient candidate combination features can be gradually eliminated, and candidate combination features that contribute more to the recognition task can be iteratively optimized. In the selection process of candidate combination features, new candidate combination features are generated through operations such as crossover and mutation, which can introduce more feature diversity. Important target feature combinations can be combined through continuous iterative optimization. The iterative optimization process can significantly reduce the computational cost of subsequent processing steps. At the same time, since the number of screened second features is small, but the effect is significant, it is possible to improve processing speed and efficiency while ensuring recognition accuracy.

[0077] In other embodiments, at least one second feature may be screened out by using an AB experiment based on the degree of influence on the recall rate and / or accuracy of the file recognition result.

[0078] Through AB experiments, the obtained second feature data set is divided into several subsets, each subset uses a different feature combination for file recognition, and then the recall rate and / or precision rate of each subset are compared to screen out the best combination of features in one or more subsets as the second feature.

[0079] In the disclosed embodiment, the AB experiment can be used to quantitatively evaluate the impact of each feature combination on the recall rate and / or accuracy of the file recognition result. By testing multiple feature combinations at the same time and comparing their effects, the best combination can be selected from multiple candidate feature combinations, thereby improving the efficiency of subsequent file recognition using the combined feature.

[0080] In the disclosed embodiment, in order to improve the accuracy and stability of the file recognition result, a feature combination rule may be determined based on the combination of at least one second feature, and the feature combination rule is used to combine the third features extracted from the file to be recognized.

[0081] The combination of the second features that will be screened out is used as a standard combination for identifying whether a file is generated by artificial intelligence technology, and is used in subsequent feature combinations for identifying files.

[0082] In the disclosed embodiment, the determined feature combination rule represents an effective feature combination rule that can more accurately identify the relevant features of the artificial intelligence generated file, which can improve the accuracy of recognition. In addition, the determined feature combination rule does not need to re-explore the feature combination every time, which can save time and computing resources and improve the overall recognition efficiency.

[0083] Finally, a known feature set is constructed by using the first feature extracted from the file template based on the aforementioned method and the second feature consisting of the combined features. It should be noted that even if there are multiple file templates of the same target type, the first feature and the second feature can be extracted from these file templates respectively, and the first features and the second features of all file templates are summarized in the known feature set, so as to facilitate feature matching with the file to be identified through the known feature set, thereby identifying the file or file page generated based on artificial intelligence technology.

[0084] Therefore, based on the same technical concept, the embodiment of the present disclosure also provides a file identification method, such as Figure 3 As shown, the following steps may be included:

[0085] S301, extracting a plurality of third features from a page to be identified of a file to be identified.

[0086] S302, combining multiple third features to obtain at least one fourth feature.

[0087] During implementation, the third characteristics can be randomly combined to obtain at least one fourth characteristic.

[0088] In order to improve the accuracy and stability of recognition, in the embodiment of the present disclosure, the third features can be combined based on the feature combination rules described above to obtain the fourth feature.

[0089] S303, matching the plurality of third features and at least one fourth feature with a known feature set of a file template required for generating a file type of a to-be-identified file based on artificial intelligence technology, to obtain a feature matching degree.

[0090] The extracted multiple third features and the at least one fourth feature obtained by combining are matched with the known feature set of the file template required by the file type generated in advance based on artificial intelligence technology. By comparing the similarity between the features of the file to be identified and the template features, the similarity is used as the feature matching degree.

[0091] S304: When the feature matching degree meets the preset conditions, determine that the page to be identified is generated based on artificial intelligence technology.

[0092] Whether the feature matching degree meets the preset conditions is used to determine whether the page to be identified is generated based on artificial intelligence technology. If the feature matching degree meets the preset conditions, it can be determined that the page to be identified is generated by artificial intelligence technology.

[0093] In the disclosed embodiment, by extracting and combining multiple third features, richer features to be matched can be formed, and these features to be matched can more comprehensively reflect the characteristics of the page to be identified. By matching these features with the known feature set of files generated based on artificial intelligence technology, it is possible to more accurately determine whether the file is generated by artificial intelligence technology. Taking into account the various variations and disguises that may exist in the file, by combining multiple features and performing complex matching logic, file identification in different situations can be handled. The setting of preset conditions also allows the sensitivity of recognition to be adjusted according to actual needs, so as to achieve the best recognition effect in different scenarios.

[0094] In the disclosed embodiment, similar to the first feature, the third feature is used to describe at least one of the typeset features of multiple page elements in the page to be identified, the relative features between different page elements, and the relative features between the same type of page elements.

[0095] The layout features and relative features of the page elements in the third feature are the same as the layout features and relative features of the template elements in the first feature. The difference is that the first feature is extracted from the file template, and the third feature is extracted from the page to be identified. Therefore, the specific content examples and extraction methods of the third feature are not repeated here.

[0096] In the disclosed embodiment, by analyzing the typeset features of multiple page elements in the page to be identified, the relative features between different page elements, and the relative features between similar page elements, it is possible to better understand the structure and layout of the page, determine whether the file is generated by artificial intelligence technology, and thus improve recognition efficiency.

[0097] In the disclosed embodiment, the feature matching degree includes a first sub-matching degree and a second sub-matching degree; the first sub-matching degree is determined based on the matching result between multiple third features and a first feature in a known feature set corresponding to the multiple third features; the second sub-matching degree is determined based on the matching result between at least one fourth feature and a second feature in a known feature set corresponding to at least one fourth feature; when the feature matching degree satisfies a preset condition, it is determined that the page to be identified is generated based on artificial intelligence technology, that is, when the first sub-matching degree is greater than a first matching threshold, and / or when the second sub-matching degree is greater than a second matching threshold, it is determined that the page to be identified is generated based on artificial intelligence technology.

[0098] During implementation, the third feature can be used as the first-level matching feature through the multi-layer feature matching method, and the first matching threshold can be set according to the a posteriori performance. For example, if 95 of the 100 third features are matched to generate a PPT, the first matching threshold is 95%. The fourth feature can be used as the second-level matching feature. Similar to the third feature strategy, multiple fourth feature matches can set the second matching threshold according to the a posteriori performance.

[0099] When either the first sub-matching degree is greater than the first matching threshold or the second sub-matching degree is greater than the second matching threshold, or both conditions are met, it is determined that the page to be identified is generated based on artificial intelligence technology.

[0100] In the disclosed embodiment, the first sub-matching degree and the second sub-matching degree can be matched for different feature types respectively, so as to adapt to different types of pages to be identified, which helps to improve the accuracy of identification and reduce the cases of misjudgment and missed identification. By comprehensively considering the matching results of the first sub-matching degree and / or the second sub-matching degree, it is possible to more comprehensively evaluate whether the page to be identified is generated based on artificial intelligence technology. By setting a matching threshold, pages that may be generated based on artificial intelligence technology can be quickly screened out, thereby improving the recognition efficiency.

[0101] In the embodiment of the present disclosure, when the fourth feature is obtained by randomly combining the third features, the first features corresponding to the third features in the known feature set may also be randomly combined in the same manner as the random combination of the third features to obtain a second feature matching the fourth feature.

[0102] Of course, in the embodiment of the present disclosure, multiple third features can also be combined according to feature combination rules to obtain at least one fourth feature; the feature combination rule is determined based on at least one second feature screened from the file template.

[0103] That is, when the aforementioned feature combination rule is used for combination, the third features are combined according to the feature combination method to obtain the fourth feature. The corresponding first features are also combined according to the feature combination method to obtain the second feature matching the fourth feature.

[0104] In the disclosed embodiment, multiple third features are combined into a fourth feature through feature combination rules, which can better reflect the essential characteristics of the files generated by artificial intelligence, thereby reducing misjudgments during file recognition and improving the accuracy of file recognition results.

[0105] In the embodiment of the present disclosure, when the content of a preset proportion of pages in the file to be identified is determined to be generated based on artificial intelligence technology, it can be determined that the file to be identified is generated based on artificial intelligence technology.

[0106] For example, the file to be identified has m pages and the preset ratio is p. When p*m pages are determined to be generated based on artificial intelligence technology, it is determined that the file to be identified is generated based on artificial intelligence technology.

[0107] In addition, in order to avoid missed detection and improve the recall rate, in the embodiment of the present disclosure, when at least one page of content in the file to be identified is determined to be generated based on artificial intelligence technology, it is determined that the file to be identified is generated based on artificial intelligence technology.

[0108] In the disclosed embodiment, by checking at least one page of content, the risk of misjudging a non-AI-generated file as an AI-generated file can be reduced to a certain extent, which helps to improve the recall rate of recognition and facilitates downstream businesses to execute corresponding processes based on the recognition results, so as to improve the accuracy and rationality of subsequent processes.

[0109] In summary, the file processing method provided by the embodiment of the present disclosure can screen out the first feature and feature combination rules for subsequent file identification by maximizing the objective function based on a genetic algorithm or by an AB experiment. The file identification method provided by the embodiment of the present disclosure can extract the third feature in the file page and the fourth feature composed of the third feature based on the combination rule obtained by the file processing method, obtain the matching result by comparing the first feature and the third feature, the second feature and the fourth feature, and determine whether the file is a file generated by artificial intelligence technology based on the matching result. Taking the identification of whether a PPT is generated by artificial intelligence technology as an example, the specific identification process is as follows: Figure 4 As shown, including:

[0110] S401, inputting a PPT file template, and converting the PPT file template into an editable json (JavaScript Object Notation, a lightweight data format) format.

[0111] S402, traversing the content of each page in the PPT file template.

[0112] S403, extracting weak features (i.e., first features) from the PPT file template. For example, including: number of graphics, size ratio relationship between various graphic types, concentric circle relationship, transparency, alignment relationship, relative position relationship, etc. An example is as follows:

[0113] weak_feature1 = number of neatly arranged small circles 20;

[0114] weak_feature2 = circle 1 and circle 2 are concentric circles;

[0115] weak_feature3 = the transparency of circle 2 is 0.75;

[0116] weak_feature4=The ratio of rounded rectangle 1 to the ppt page is 0.2:1;

[0117] weak_feature5=Rounded rectangle 1 and rounded rectangle 2 are horizontally aligned; ......

[0119] S404, combining weak features into strong features (i.e., second features). The weak features extracted in S403 are combined into strong features by random combination, for example:

[0120] strong_feature1=weak_feature1&weak_feature7;

[0121] strong_feature2=weak_feature2&weak_feature3&weak_feature5;

[0122] strong_feature3=weak_feature3&weak_feature5&weak_feature8; ......

[0124] S405, constructing a known feature set through weak features and strong features.

[0125] S406, extract weak features from the page to be identified of the PPT file to be identified and combine them into strong features. Then match the weak features and strong features with the known feature set constructed above to obtain feature matching degree. When the feature matching degree meets the preset conditions, it is determined that the page to be identified of the PPT file to be identified is generated based on artificial intelligence technology.

[0126] The page to be identified is any page in the PPT template to be identified.

[0127] S407, judging whether the traversal of the pages of the PPT to be identified is completed. If the traversal is not completed, returning to S406; if the traversal is completed, continuing to S408.

[0128] S408, comprehensively analyzing the multi-page determination results of the PPT to be identified, and determining whether the overall PPT file template is generated by artificial intelligence technology.

[0129] Based on the same technical concept, the present disclosure also provides a file processing device 500, such as Figure 5 As shown, including:

[0130] A first extraction module 501 is used to extract a plurality of first features from a file template required for generating a target type file based on artificial intelligence technology;

[0131] The screening module 502 is used to combine multiple first features to screen at least one second feature that can identify a file page generated based on artificial intelligence technology.

[0132] In some embodiments, the screening module comprises:

[0133] The screening subunit is used to screen out at least one second feature based on the recall rate and / or accuracy rate of the file recognition result.

[0134] In some embodiments, the screening subunit is further used to:

[0135] Taking maximizing the objective function as a criterion, multiple first features are combined based on a genetic algorithm to obtain at least one second feature; the objective function includes a recall rate and / or an accuracy rate of a file recognition result.

[0136] In some embodiments, the screening subunit is specifically used for:

[0137] Randomly combining multiple first features to obtain multiple candidate combination features;

[0138] Based on each candidate combination feature, the function value of the objective function is determined respectively;

[0139] According to the size of the function value, multiple seed combination features are screened out from multiple candidate combination features;

[0140] A crossover and mutation operation is performed on the multiple seed combination features to obtain multiple new candidate combination features, and the step of determining the function value of the objective function based on each candidate combination feature is returned to execute until at least one second feature that maximizes the objective function is screened out.

[0141] In some embodiments, the screening subunit is further used to:

[0142] Based on the influence degree on the recall rate and / or precision rate of the file recognition result, at least one second feature is screened out by adopting an AB experiment method.

[0143] In some embodiments, it also includes:

[0144] Based on the combination mode of at least one second feature, a feature combination rule is determined, and the feature combination rule is used to combine the third features extracted from the file to be identified.

[0145] In some embodiments, the first feature is used to describe at least one of the typeset features of each of the multiple template elements in the document template, the relative features between different template elements, and the relative features between the same type of template elements.

[0146] Based on the same technical concept, the present disclosure also provides a file identification device 600, such as Figure 6 As shown, including:

[0147] The second extraction module 601 is used to extract a plurality of third features from the to-be-identified page of the to-be-identified file;

[0148] A combining module 602, used to combine multiple third features to obtain at least one fourth feature;

[0149] A matching module 603 is used to match the plurality of third features and at least one fourth feature with a known feature set of a file template required for generating a file type of a to-be-identified file based on artificial intelligence technology, to obtain a feature matching degree;

[0150] The determination module 603 is used to determine that the page to be identified is generated based on artificial intelligence technology when the feature matching degree meets the preset conditions.

[0151] In some embodiments, the third feature is used to describe at least one of the layout features of multiple page elements in the page to be identified, the relative features between different page elements, and the relative features between the same type of page elements.

[0152] In some embodiments, the feature matching degree includes a first sub-matching degree and a second sub-matching degree; the first sub-matching degree is determined based on the matching result between the plurality of third features and the first features in the known feature set corresponding to the plurality of third features; the second sub-matching degree is determined based on the matching result between the at least one fourth feature and the second feature in the known feature set corresponding to the at least one second feature;

[0153] Identify modules, specifically for:

[0154] When the first sub-matching degree is greater than the first matching threshold, and / or the second sub-matching degree is greater than the second matching threshold, it is determined that the page to be identified is generated based on artificial intelligence technology.

[0155] In some embodiments, the combined module comprises:

[0156] According to the feature combination rule, multiple third features are combined to obtain at least one fourth feature; the feature combination rule is determined based on at least one second feature screened from the file template.

[0157] In some embodiments, the determining module is further configured to:

[0158] When at least one page of content in the file to be identified is determined to be generated based on artificial intelligence technology, it is determined that the file to be identified is generated based on artificial intelligence technology.

[0159] For the description of specific functions and examples of each module and submodule of the device in the embodiment of the present disclosure, reference can be made to the relevant description of the corresponding steps in the above method embodiment, which will not be repeated here.

[0160] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0161] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0162] Figure 7 A schematic block diagram of an example electronic device 700 that can be used to implement an embodiment of the present disclosure is shown. 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 can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0163] like Figure 7 As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0164] A number of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0165] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as a file processing method and / or a file identification method. For example, in some embodiments, the file processing method and / or the file identification method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the file processing method and / or the file identification method described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to execute the file processing method and / or the file identification method in any other appropriate manner (eg, by means of firmware).

[0166] Various implementations 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 chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0167] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0168] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0169] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the 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 acoustic input, voice input, or tactile input).

[0170] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0171] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0172] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0173] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A file processing method, comprising: Extracting a plurality of first features from a file template required for generating a target type file based on artificial intelligence technology; The multiple first features are combined to filter at least one second feature that can identify a file page generated based on artificial intelligence technology.

2. The method according to claim 1, wherein: The combining of the plurality of first features to screen at least one second feature capable of identifying a file page generated based on artificial intelligence technology includes: At least one second feature is selected based on the recall rate and / or the accuracy rate of the document recognition result.

3. The method according to claim 2, wherein: The step of selecting at least one second feature based on the recall rate and / or accuracy rate of the file recognition result includes: The plurality of first features are combined based on a genetic algorithm with the maximization of an objective function as a criterion to obtain the at least one second feature; the objective function includes a recall rate and / or an accuracy rate of a file recognition result.

4. The method according to claim 3, wherein: The step of combining the plurality of first features based on a genetic algorithm to obtain the at least one second feature by taking maximizing the objective function as a criterion includes: Randomly combining the multiple first features to obtain multiple candidate combination features; Based on each candidate combination feature, respectively determine the function value of the objective function; According to the magnitude of the function value, a plurality of seed combination features are selected from the plurality of candidate combination features; Perform crossover and mutation operations on the multiple seed combination features to obtain multiple new candidate combination features, and return to execute the step of determining the function value of the objective function based on each candidate combination feature, until the at least one second feature that maximizes the objective function is screened out.

5. The method according to claim 2, wherein: The step of selecting at least one second feature based on the recall rate and / or accuracy rate of the file recognition result includes: Based on the influence degree on the recall rate and / or the accuracy rate of the file recognition result, the at least one second feature is screened out by adopting an AB experiment method.

6. The method according to any one of claims 1 to 5, further comprising: Based on the combination mode of the at least one second feature, a feature combination rule is determined, and the feature combination rule is used to combine the third features extracted from the file to be identified.

7. The method according to any one of claims 1 to 5, wherein: The first feature is used to describe at least one of the typeset features of each of the multiple template elements in the file template, the relative features between different template elements, and the relative features between the same type of template elements.

8. A file identification method, comprising: Extracting a plurality of third features from the to-be-identified pages of the to-be-identified file; Combining the plurality of third features to obtain at least one fourth feature; Matching the plurality of third features and the at least one fourth feature with a known feature set of a file template required for generating the file type of the to-be-identified file based on artificial intelligence technology to obtain a feature matching degree; When the feature matching degree meets a preset condition, it is determined that the page to be identified is generated based on artificial intelligence technology.

9. The method according to claim 8, wherein: The third feature is used to describe at least one of the typeset features of each of the multiple page elements in the page to be identified, the relative features between different page elements, and the relative features between the same type of page elements.

10. The method according to claim 8, wherein: The feature matching degree includes a first sub-matching degree and a second sub-matching degree; the first sub-matching degree is determined based on the matching result between the plurality of third features and the first features in the known feature set corresponding to the plurality of third features; the second sub-matching degree is determined based on the matching result between the at least one fourth feature and the second feature in the known feature set corresponding to the at least one fourth feature; When the feature matching degree satisfies a preset condition, determining that the page to be identified is generated based on artificial intelligence technology includes: When the first sub-matching degree is greater than a first matching threshold, and / or the second sub-matching degree is greater than a second matching threshold, it is determined that the page to be identified is generated based on artificial intelligence technology.

11. The method according to claim 8, wherein: Combining the plurality of third features to obtain at least one fourth feature comprises: The multiple third features are combined according to a feature combination rule to obtain the at least one fourth feature; the feature combination rule is determined based on at least one second feature screened out from the file template.

12. The method according to any one of claims 8 to 11, further comprising: In the case where at least one page of content in the file to be identified is determined to be generated based on artificial intelligence technology, it is determined that the file to be identified is generated based on artificial intelligence technology.

13. A file processing device, comprising: A first extraction module, used to extract a plurality of first features from a file template required for generating a target type file based on artificial intelligence technology; The screening module is used to combine the multiple first features to screen at least one second feature that can identify the file page generated based on artificial intelligence technology.

14. The device according to claim 13, wherein: The screening module comprises: The screening subunit is used to screen out at least one second feature based on the recall rate and / or accuracy rate of the file recognition result.

15. The device according to claim 14, wherein: The screening subunit is specifically used for: The plurality of first features are combined based on a genetic algorithm with the maximization of an objective function as a criterion to obtain the at least one second feature; the objective function includes a recall rate and / or an accuracy rate of a file recognition result.

16. The device according to claim 15, wherein: The screening subunit is specifically used for: Randomly combining the multiple first features to obtain multiple candidate combination features; Based on each candidate combination feature, respectively determine the function value of the objective function; According to the magnitude of the function value, a plurality of seed combination features are selected from the plurality of candidate combination features; Perform crossover and mutation operations on the multiple seed combination features to obtain multiple new candidate combination features, and return to execute the step of determining the function value of the objective function based on each candidate combination feature, until the at least one second feature that maximizes the objective function is screened out.

17. The device according to claim 14, wherein: The screening subunit is specifically used for: Based on the influence degree on the recall rate and / or the accuracy rate of the file recognition result, the at least one second feature is screened out by adopting an AB experiment method.

18. The apparatus according to any one of claims 13 to 17, further comprising: Based on the combination mode of the at least one second feature, a feature combination rule is determined, and the feature combination rule is used to combine the third features extracted from the file to be identified.

19. The device according to any one of claims 13 to 17, wherein: The first feature is used to describe at least one of the typeset features of each of the multiple template elements in the file template, the relative features between different template elements, and the relative features between the same type of template elements.

20. A file recognition device, comprising: A second extraction module is used to extract a plurality of third features from the to-be-identified page of the to-be-identified file; A combining module, used for combining the plurality of third features to obtain at least one fourth feature; a matching module, configured to match the plurality of third features and the at least one fourth feature with a known feature set of a file template required for generating the file type of the to-be-identified file based on artificial intelligence technology, to obtain a feature matching degree; The determination module is used to determine that the page to be identified is generated based on artificial intelligence technology when the feature matching degree meets a preset condition.

21. The device according to claim 20, wherein: The third feature is used to describe at least one of the typeset features of each of the multiple page elements in the page to be identified, the relative features between different page elements, and the relative features between the same type of page elements.

22. The device according to claim 20, wherein: The feature matching degree includes a first sub-matching degree and a second sub-matching degree; the first sub-matching degree is determined based on the matching result between the plurality of third features and the first features in the known feature set corresponding to the plurality of third features; the second sub-matching degree is determined based on the matching result between the at least one fourth feature and the second feature in the known feature set corresponding to the at least one second feature; The determination module is specifically used for: When the first sub-matching degree is greater than a first matching threshold, and / or the second sub-matching degree is greater than a second matching threshold, it is determined that the page to be identified is generated based on artificial intelligence technology.

23. The device according to claim 20, wherein: The combined module comprises: The multiple third features are combined according to a feature combination rule to obtain the at least one fourth feature; the feature combination rule is determined based on at least one second feature screened out from the file template.

24. The apparatus according to any one of claims 20 to 23, wherein the determining module is further configured to: In the case where at least one page of content in the file to be identified is determined to be generated based on artificial intelligence technology, it is determined that the file to be identified is generated based on artificial intelligence technology.

25. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 12.

26. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-12.

27. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 12.

Citation Information

Patent Citations

  • Text recognition method and device, processor and electronic equipment

    CN116955624A

  • Paper detection method and device, storage medium and terminal

    CN117033555A

  • Method, medium and system for identifying AI generation text contained in test paper

    CN117113977A

  • Large language model generation text detection method based on ensemble learning

    CN117216687A

  • Malicious file detection method and device, equipment and storage medium

    CN117370966A