A method and system for automatic conversion of electronic files based on digitization

Abstract

The present invention discloses a method and system for automatic conversion of electronic archives based on digitization, and relates to the field of electronic archives. The method comprises the following steps: using media type recognition technology to judge the type of input archive data, and distinguishing the input archive data into image archive files and audio archive files; using optical character recognition technology to digitally recognize and convert the image archive files to obtain an initial text file; combining the audio archive files with the initial text files to construct a voice correction model; calculating the matching degree through the voice correction model, and correcting the initial text file according to the matching degree; organizing and archiving the corrected text file with the image archive files and the audio archive files to complete the electronic archive conversion. The present invention can assist in completing the conversion of electronic archives when the electronic archive input information is incomplete through voice recognition technology, thereby improving the efficiency and quality of archive conversion.

Description

Technical Field

[0001] The present invention relates to the field of electronic archives, and in particular to an automatic conversion method and system for electronic archives based on digitization. Background Art

[0002] With the development of society, the application scope of electronic archives is constantly expanding. When it comes to special groups and special scenarios, the problems of low efficiency and uneven conversion quality of traditional electronic archive conversion methods have gradually become prominent: First, due to the continuous increase in the proportion of the elderly population in my country in recent years, a large number of elderly people have participated in providing archival information. However, the elderly are unfamiliar with emerging technologies, and the information they enter into electronic archives using smart devices is incomplete and the error rate is high. They also need to rely on voice descriptions for correction. However, the existing electronic archive conversion system cannot effectively use voice for correction, resulting in the inability to effectively digitize the important archival information provided by the elderly, and often requires manual correction, which increases the cost of digitizing electronic archives. In addition, a large amount of audio materials are also generated while recording archives in important scenarios such as courts, meetings, and press releases. These audio materials can serve as important supplementary information for electronic archives. However, most of the audio content is not converted and utilized but is directly archived, resulting in these important voice information being unable to supplement the digitized archives, and also losing the opportunity to assist in improving the quality of electronic archive conversion.

[0003] In order to solve the above problems, it is urgently necessary to study a new method that can realize automatic and efficient conversion of electronic archives. This method should be able to assist in the conversion of electronic archives through voice recognition technology when the electronic archive entry information is incomplete, and realize the accurate digitization of the voice description of the elderly. It is also necessary to be able to assist in extracting key information from audio materials and correct and supplement electronic archives. Building such an automated electronic archive conversion system will greatly improve the efficiency and quality of archive conversion and has important social value.

[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0005] In view of the problems in the related art, the present invention proposes a method and system for automatic conversion of electronic files based on digitization to solve the above technical problems existing in the prior art.

[0006] To this end, the specific technical solution adopted by the present invention is as follows:

[0007] According to one aspect of the present invention, a method for automatic conversion of electronic archives based on digitization is provided, and the method for automatic conversion of electronic archives based on digitization comprises the following steps:

[0008] S1, using media type recognition technology to determine the type of input archive data, and distinguish the input archive data into image archive files and audio archive files;

[0009] S2, using optical character recognition technology to digitally identify and convert the image archive file to obtain an initial text file;

[0010] S3, combining the audio archive file and the initial text file to build a speech correction model;

[0011] S4, calculating the matching degree through the speech correction model, and correcting the initial text file according to the matching degree;

[0012] S5. Organize and archive the revised text files, image archive files, and audio archive files to complete the electronic archive conversion.

[0013] Furthermore, using the media type recognition technology to determine the type of the input archive data and distinguishing the input archive data into image archive files and audio archive files includes the following steps:

[0014] S11, obtaining the file extension and file header information of the input archive data;

[0015] S12, establishing a media type identification rule base based on the file extension and file header information;

[0016] S13, defining the image archive file type and the audio archive file type in the media type identification rule library;

[0017] S14, comparing the file extension and file header information of the input archive data with the image archive file type and the audio archive file type defined in the media type identification rule base;

[0018] S15. Based on the comparison result, determine whether the input data is an image archive file or an audio archive file, and complete the file type identification of the input archive data.

[0019] Furthermore, the optical character recognition technology is used to digitally identify and convert the image archive file to obtain the initial text file, which includes the following steps:

[0020] S21, digitally scanning the image archive file using a scanning device to obtain original image data in a predetermined format;

[0021] S22, using optical character recognition technology to perform preliminary recognition and conversion on the original image data, and convert the text in the image into a draft text;

[0022] S23, formatting the draft text according to the predetermined standard format requirements;

[0023] S24, saving the formatted text data as a standard text file as an initial text file.

[0024] Furthermore, using optical character recognition technology to preliminarily recognize and convert the original image data, converting the text in the image into a draft text includes the following steps:

[0025] S221, graying and noise reduction preprocessing the original image data;

[0026] S222, analyzing the text layout, paragraphs and titles in the image to obtain image structure information;

[0027] S223, determining text lines and text positions, and segmenting and extracting the image;

[0028] S224, using feature extraction and classifier to identify the extracted character image to obtain the corresponding character;

[0029] S225, combining characters into words, restoring punctuation marks, and obtaining a draft text by document typesetting.

[0030] Furthermore, combining the audio archive file with the initial text file, constructing a speech correction model includes the following steps:

[0031] S31, performing speech recognition on the audio archive file to obtain a speech recognition text;

[0032] S32, comparing the speech recognition text with the initial text file to find the text differences between the two;

[0033] S33, by analyzing the text differences, marking the corresponding voice segments in the audio, and constructing a voice error correction data set;

[0034] S34, constructing a speech correction model, and using the speech error correction data set to train the speech correction model;

[0035] S35, the voice clips, text differences, and image contents are displayed in association, and the error correction results are fed back through manual verification to optimize the voice correction model.

[0036] Furthermore, by analyzing the text differences and marking the corresponding speech segments in the audio, constructing a speech error correction dataset includes the following steps:

[0037] S331, finding the text difference between the speech recognition text and the initial text file through a string comparison algorithm, and locating the wrong words;

[0038] S332, marking incorrect words in the initial text file;

[0039] S333, determining the time period of the incorrect word in the audio according to the alignment relationship between the speech recognition text and the annotated initial text file;

[0040] S334. Extract audio clips of incorrect words from the audio archive file to form an audio error correction data set.

[0041] Furthermore, the matching degree is calculated by the speech correction model, and the initial text file is corrected according to the matching degree, including the following steps:

[0042] S41, inputting the speech recognition text and the initial text file into the speech correction model;

[0043] S42, comparing the speech recognition text with the initial text file through the speech correction model, and calculating the text fitness;

[0044] S43, marking incorrect words in the initial text file according to the text fitness, and generating audio clips of these incorrect words through the speech correction model;

[0045] S44, the speech correction model corrects the incorrect vocabulary and outputs a recognition result consistent with the audio clip;

[0046] S45, replacing the recognition result to the corresponding error position in the initial text file to generate a corrected text file.

[0047] Furthermore, the speech recognition text is compared with the initial text file through the speech correction model, and the text fitness is calculated, including the following steps:

[0048] S421, performing denoising preprocessing on the speech recognition text and the initial text file;

[0049] S421, extracting features of entity categories from the preprocessed text, and performing feature analysis on each entity category;

[0050] S423, calculate the fitness of each pair of entity categories, and integrate them to obtain the overall text fitness. ;

[0051] S424: Set a fitness threshold, and evaluate and adjust the calculated text fitness.

[0052] Furthermore, the formula for calculating text fitness is:

[0053]

[0054] In the formula, SYD(J 1 , J 2 ) represents the overall text adaptability of the speech recognition text to the initial text file;

[0055] J1 A text feature set representing the speech recognition text;

[0056] J 2 A text feature set representing the composition of the initial text file;

[0057] i represents the i-th wrong word in the text feature set;

[0058] K represents the total number of incorrect words in the text feature set;

[0059] h 1i Represents speech recognition text J 1 The corresponding words for the wrong words in the sentence;

[0060] h 2i Represents the initial text file J 2 Wrong vocabulary in .

[0061] According to another aspect of the present invention, there is also provided an automatic conversion system for electronic archives based on digitization, the automatic conversion system for electronic archives based on digitization comprising:

[0062] A judgment and distinction module, used for judging the type of input archive data by using media type recognition technology, and distinguishing the input archive data into image archive files and audio archive files;

[0063] An image recognition module is used to digitally recognize and convert image archive files using optical character recognition technology to obtain an initial text file;

[0064] A model building module, used to combine the audio archive file with the initial text file to build a speech correction model;

[0065] A text correction module is used to calculate the matching degree through the speech correction model and correct the initial text file according to the matching degree;

[0066] The file conversion module is used to organize and archive the revised text files, image file files and audio file files to complete the electronic file conversion.

[0067] The beneficial effects of the present invention are:

[0068] (1) Through optical character recognition technology and audio content extraction technology, the present invention realizes the digital utilization of elderly people's voice descriptions and important scene audio materials, solves the problem that traditional methods cannot process voice input and a large amount of audio content, and greatly improves the quality and efficiency of file conversion; at the same time, the present invention can reduce the operational difficulty of the elderly participating in electronic file conversion, reduce the error rate, and can also assist in correcting and supplementing the content of electronic files, so as to realize the effective utilization of audio materials; in addition, the present invention also promotes the development of electronic file conversion towards automation and intelligence, making digital electronic file conversion more efficient and accurate, and providing users with better quality and more convenient electronic file conversion services.

[0069] (2) In view of the difficulty faced by the elderly in participating in the electronic archive conversion process, the present invention can directly recognize the voice description provided by the elderly through voice recognition technology and generate electronic text, thereby solving the problem of difficulty in participation due to the elderly's unfamiliarity with new technologies. At the same time, the present invention can guide the elderly to describe information as required through voice interaction and other forms, thereby improving the effectiveness of voice description, lowering the threshold for the elderly in the electronic archive conversion process, and making it easier for the elderly to participate in the electronic archive conversion process. In addition, voice recognition can correct and supplement incomplete and erroneous information input by the elderly, thereby reducing manual intervention and lowering conversion costs.

[0070] (3) In order to address the problem that the massive amount of audio data generated in important scenes is not fully utilized, the present invention enriches the information content of electronic archives and improves the quality of archive conversion through voice recognition and key information extraction technology, solves the problem of resource waste caused by the isolated archiving of audio data, and greatly expands the information space of electronic archives; at the same time, it can also automatically convert these audio contents and assist in correcting and supplementing electronic archives, thereby realizing the effective digital utilization of audio data. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0072] Figure 1 It is a flowchart of a method for automatic conversion of electronic archives based on digitization according to an embodiment of the present invention;

[0073] Figure 2 The present invention is a block diagram of a digital electronic archive automatic conversion system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0074] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0075] According to an embodiment of the present invention, a method and system for automatic conversion of electronic files based on digitization are provided.

[0076] The present invention is further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, the method for automatic conversion of electronic archives based on digitization according to an embodiment of the present invention includes the following steps:

[0077] S1, using media type recognition technology to determine the type of input archive data, and distinguish the input archive data into image archive files and audio archive files;

[0078] S2, using optical character recognition technology to digitally identify and convert the image archive file to obtain an initial text file;

[0079] S3, combining the audio archive file and the initial text file to build a speech correction model;

[0080] S4, calculating the matching degree through the speech correction model, and correcting the initial text file according to the matching degree;

[0081] S5. Organize and archive the revised text files, image archive files, and audio archive files to complete the electronic archive conversion.

[0082] In one embodiment, using media type recognition technology to determine the type of input archive data and distinguishing the input archive data into image archive files and audio archive files includes the following steps:

[0083] S11, obtaining the file extension and file header information of the input archive data;

[0084] S12, establishing a media type identification rule base based on the file extension and file header information;

[0085] S13, defining the image archive file type and the audio archive file type in the media type identification rule library;

[0086] S14, comparing the file extension and file header information of the input archive data with the image archive file type and the audio archive file type defined in the media type identification rule base;

[0087] S15. Based on the comparison result, determine whether the input data is an image archive file or an audio archive file, and complete the file type identification of the input archive data.

[0088] In one embodiment, the process of digitally identifying and converting the image archive file using optical character recognition technology to obtain the initial text file includes the following steps:

[0089] S21, digitally scanning the image archive file using a scanning device to obtain original image data in a predetermined format;

[0090] S22, using optical character recognition technology to perform preliminary recognition and conversion on the original image data, and convert the text in the image into a draft text;

[0091] Specifically, using optical character recognition technology to preliminarily recognize and convert the original image data, and converting the text in the image into draft text includes the following steps:

[0092] S221, graying and noise reduction preprocessing the original image data;

[0093] S222, analyzing the text layout, paragraphs and titles in the image to obtain image structure information;

[0094] S223, determining text lines and text positions, and segmenting and extracting the image;

[0095] S224, using feature extraction and classifier to identify the extracted character image to obtain the corresponding character;

[0096] S225, combining characters into words, restoring punctuation marks, and obtaining a draft text by document typesetting.

[0097] S23, formatting the draft text according to the predetermined standard format requirements;

[0098] S24, saving the formatted text data as a standard text file as an initial text file.

[0099] In one embodiment, combining the audio archive file with the initial text file to construct a speech correction model includes the following steps:

[0100] S31, performing speech recognition on the audio archive file to obtain a speech recognition text;

[0101] S32, comparing the speech recognition text with the initial text file to find the text differences between the two;

[0102] S33, by analyzing the text differences, marking the corresponding voice segments in the audio, and constructing a voice error correction data set;

[0103] Specifically, by analyzing the text differences and marking the corresponding speech segments in the audio, constructing a speech error correction dataset includes the following steps:

[0104] S331, finding the text difference between the speech recognition text and the initial text file through a string comparison algorithm, and locating the wrong words;

[0105] S332, marking incorrect words in the initial text file;

[0106] S333, determining the time period of the incorrect word in the audio according to the alignment relationship between the speech recognition text and the annotated initial text file;

[0107] S334. Extract audio clips of incorrect words from the audio archive file to form an audio error correction data set.

[0108] S34, constructing a speech correction model, and using the speech error correction data set to train the speech correction model;

[0109] S35, the voice clips, text differences, and image contents are displayed in association, and the error correction results are fed back through manual verification to optimize the voice correction model.

[0110] In one embodiment, calculating the matching degree through the speech correction model and correcting the initial text file according to the matching degree includes the following steps:

[0111] S41, inputting the speech recognition text and the initial text file into the speech correction model;

[0112] S42, the fitness correction unit compares the speech recognition text with the initial text file and calculates the text fitness;

[0113] Specifically, comparing the speech recognition text with the initial text file through the speech correction model and calculating the text fitness includes the following steps:

[0114] S421, performing denoising preprocessing on the speech recognition text and the initial text file;

[0115] S421, extracting features of entity categories from the preprocessed text, and performing feature analysis on each entity category;

[0116] S423, calculate the fitness of each pair of entity categories, and integrate them to obtain the overall text fitness. ;

[0117] S424: Set a fitness threshold, and evaluate and adjust the calculated text fitness.

[0118] Specifically, the formula for calculating text fitness is:

[0119]

[0120] In the formula, SYD(J 1 , J 2 ) represents the overall text adaptability of the speech recognition text to the initial text file;

[0121] J 1 A text feature set representing the speech recognition text;

[0122] J 2 A text feature set representing the composition of the initial text file;

[0123] i represents the i-th wrong word in the text feature set;

[0124] K represents the total number of incorrect words in the text feature set;

[0125] h 1i Represents speech recognition text J 1 The corresponding words for the wrong words in the sentence;

[0126] h 2i Represents the initial text file J 2 Wrong vocabulary in .

[0127] S43, marking incorrect words in the initial text file according to the text fitness, and generating audio clips of these incorrect words through a speech recognition unit;

[0128] S44, the speech recognition unit corrects the incorrect words and outputs a recognition result consistent with the audio clip;

[0129] S45, replacing the recognition result to the corresponding error position in the initial text file to generate a corrected text file.

[0130] Specifically, by calculating the fitness of all speech recognition texts and the erroneous words in the initial text file, a larger fitness indicates a smaller corresponding vector angle and a more similarity between the two. A threshold is set (in the present invention, the threshold is set to 0.94), and the corresponding words with a fitness greater than the threshold can be used as the corrected text.

[0131] like Figure 2 According to another embodiment of the present invention, there is also provided a digital electronic archive automatic conversion system, the digital electronic archive automatic conversion system comprising:

[0132] A judgment and distinction module 1 is used to judge the type of input archive data by using a media type recognition technology, and to distinguish the input archive data into an image archive file and an audio archive file;

[0133] Image recognition module 2, used to digitally recognize and convert image archive files using optical character recognition technology to obtain initial text files;

[0134] A model building module 3 is used to combine the audio archive file with the initial text file to build a speech correction model;

[0135] The text correction module 4 is used to calculate the matching degree through the speech correction model and correct the initial text file according to the matching degree;

[0136] The file conversion module 5 is used to organize and archive the revised text file, image file and audio file to complete the electronic file conversion.

[0137] In order to facilitate understanding of the above technical solutions of the present invention, the working principle or operation mode of the present invention in the actual process is described in detail below.

[0138] In actual application, taking an elderly person participating in providing historical picture archive information as an example, the above method is used to convert electronic archives. During the scanning process of the pictures, the scanning result of one picture appears blurred. Here, the optical character recognition technology makes a recognition error due to the picture quality problem. The optical character recognition technology recognition result of the picture is "My colleagues and I are working in the office", but the audio file description provided is "My classmates and I are playing on the campus". At this time, the speech correction model constructed by the present invention, combined with the audio archive file and the initial text file, finds the text difference, and corrects the initial text file according to the matching degree, and corrects the text file to "My classmates and I are playing on the campus". In this way, the speech information recognized by the audio file is used to automatically correct the recognition error of the picture description, thereby ensuring the conversion quality of the electronic archive. The entire process does not require manual word-by-word proofreading, thereby greatly improving the conversion efficiency of the electronic archive.

[0139] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for automatic conversion of electronic files based on digitization, characterized in that: The method for automatic conversion of electronic archives based on digitization comprises the following steps: S1, using media type recognition technology to determine the type of input archive data, and distinguish the input archive data into image archive files and audio archive files; S2, using optical character recognition technology to digitally identify and convert the image archive file to obtain an initial text file; S3, combining the audio archive file and the initial text file to build a speech correction model; S4, calculating the matching degree through the speech correction model, and correcting the initial text file according to the matching degree; S5, organizing and archiving the revised text file, image file and audio file to complete the electronic file conversion; The method of calculating the matching degree by the speech correction model and correcting the initial text file according to the matching degree comprises the following steps: S41, inputting the speech recognition text and the initial text file into the speech correction model; S42, comparing the speech recognition text with the initial text file through the speech correction model, and calculating the text fitness; S43, marking incorrect words in the initial text file according to the text fitness, and generating audio clips of these incorrect words through the speech correction model; S44, the speech correction model corrects the incorrect vocabulary and outputs a recognition result consistent with the audio clip; S45, replacing the recognition result into the corresponding error position in the initial text file to generate a corrected text file; The method of comparing the speech recognition text with the initial text file by using the speech correction model and calculating the text fitness comprises the following steps: S421, performing denoising preprocessing on the speech recognition text and the initial text file; S421, extracting features of entity categories from the preprocessed text, and performing feature analysis on each entity category; S423, calculating the fitness of each pair of entity categories, and integrating them to obtain the overall text fitness; S424, setting a fitness threshold, and evaluating and adjusting the calculated text fitness; The formula for calculating text fitness is: ； In the formula, SYD(J 1 ，J 2 ) Indicates the overall text fitness of the speech recognition text and the initial text file; J 1 A text feature set representing the speech recognition text; J 2 A text feature set representing the composition of the initial text file; i Represents the first i wrong words; K Indicates the total number of incorrect words in the text feature set; h 1i Represents speech recognition text J 1 The corresponding words for the wrong words in the sentence; h 2i Represents the initial text file J 2 Wrong vocabulary in .

2. The method for automatic conversion of electronic files based on digitization according to claim 1 is characterized in that: The method of using the media type recognition technology to determine the type of the input archive data and distinguishing the input archive data into image archive files and audio archive files includes the following steps: S11, obtaining the file extension and file header information of the input archive data; S12, establishing a media type identification rule base based on the file extension and file header information; S13, defining the image archive file type and the audio archive file type in the media type identification rule library; S14, comparing the file extension and file header information of the input archive data with the image archive file type and the audio archive file type defined in the media type identification rule base; S15. Based on the comparison result, determine whether the input data is an image archive file or an audio archive file, and complete the file type identification of the input archive data.

3. The method for automatic conversion of electronic files based on digitization according to claim 1 is characterized in that: The method of using optical character recognition technology to digitally identify and convert the image archive file to obtain the initial text file includes the following steps: S21, digitally scanning the image archive file using a scanning device to obtain original image data in a predetermined format; S22, using optical character recognition technology to perform preliminary recognition and conversion on the original image data, and convert the text in the image into a draft text; S23, formatting the draft text according to the predetermined standard format requirements; S24, saving the formatted text data as a standard text file as an initial text file.

4. The method for automatic conversion of electronic files based on digitization according to claim 3 is characterized in that: The method of using optical character recognition technology to perform preliminary recognition and conversion of the original image data to convert the text in the image into a draft text comprises the following steps: S221, graying and noise reduction preprocessing the original image data; S222, analyzing the text layout, paragraphs and titles in the image to obtain image structure information; S223, determining text lines and text positions, and segmenting and extracting the image; S224, using feature extraction and classifier to identify the extracted character image to obtain the corresponding character; S225. Combining characters into words, restoring punctuation marks, and obtaining a draft text by document typesetting.

5. The method for automatic conversion of electronic files based on digitization according to claim 1 is characterized in that: The method of combining the audio archive file with the initial text file to construct a speech correction model comprises the following steps: S31, performing speech recognition on the audio archive file to obtain a speech recognition text; S32, comparing the speech recognition text with the initial text file to find the text differences between the two; S33, by analyzing the text differences, marking the corresponding voice segments in the audio, and constructing a voice error correction data set; S34, constructing a speech correction model, and using the speech error correction data set to train the speech correction model; S35, the voice clips, text differences, and image contents are displayed in association, and the error correction results are fed back through manual verification to optimize the voice correction model.

6. The method for automatic conversion of electronic files based on digitization according to claim 5 is characterized in that: The method of analyzing text differences, marking the speech segments in the corresponding audio, and constructing a speech error correction data set includes the following steps: S331, finding the text difference between the speech recognition text and the initial text file through a string comparison algorithm, and locating the wrong words; S332, marking incorrect words in the initial text file; S333, determining the time period of the incorrect word in the audio according to the alignment relationship between the speech recognition text and the annotated initial text file; S334. Extract audio clips of incorrect words from the audio archive file to form an audio error correction data set.

7. A digital electronic file automatic conversion system, used to implement the digital electronic file automatic conversion method according to any one of claims 1 to 6, characterized in that: The digital-based electronic archive automatic conversion system includes: A judgment and distinction module, used for judging the type of input archive data by using media type recognition technology, and distinguishing the input archive data into image archive files and audio archive files; An image recognition module is used to digitally recognize and convert image archive files using optical character recognition technology to obtain an initial text file; A model building module, used to combine the audio archive file with the initial text file to build a speech correction model; A text correction module is used to calculate the matching degree through the speech correction model and correct the initial text file according to the matching degree; The file conversion module is used to organize and archive the revised text files, image file files and audio file files to complete the electronic file conversion.

Images