Literature digital processing auxiliary method and device for improving precision and efficiency

Abstract

The invention discloses a literature digital processing auxiliary method and device for improving precision and efficiency. The method comprises the following steps: carrying out labeling result preprocessing on image data labeled by a user; the image data comprises a plurality of picture words; carrying out OCR (Optical Character Recognition) on the preprocessed image data, and obtaining a corresponding recognition word for each image word; judging the wrong recognition possibility of each character based on the pixel distribution of the recognition characters and the picture characters; in response to a judgment result of possible wrong recognition, intercepting the corresponding picture word from the image data to obtain a target screenshot word; and searching a target pre-stored screenshot word from a word library based on the target screenshot word, and replacing the corresponding identification word with a word corresponding to the target pre-stored screenshot word. According to the scheme, a self error correction mechanism is designed in the process of converting the OCR into the text, the result of possible recognition errors is judged through pixel distribution, and replaceable characters with higher accuracy are searched by using the screenshot characters, so that wrong recognition characters can be obviously reduced, and the proofreading efficiency and accuracy are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of document digital processing, and in particular relates to a document digital processing auxiliary method and device for improving accuracy and efficiency. Background Art

[0002] Documents are important carriers of knowledge. Traditional documents mostly exist in physical forms such as paper. Retrieving and reading physical documents is extremely inconvenient. When faced with massive amounts of information, researchers often have to spend a lot of time and energy to browse and search, which is inefficient.

[0003] With the rapid development of computer technology, the Internet, and artificial intelligence, digital processing has provided a viable solution to these challenges. Through technologies like image scanning and OCR, paper documents can be quickly converted into digital images and then further recognized into editable text.

[0004] Currently, the digital processing of documents mainly includes data rough processing, annotation, OCR recognition, and review and correction. Data rough processing involves scanning physical documents to obtain image data. Annotation is the process of manually dividing image data into columns and marking frames. This process prepares for subsequent OCR recognition. The more standardized and accurate the frames, the more accurate the OCR recognition will be. OCR recognition is to recognize the image words in the image data and convert them into editable text according to the annotation results. The review and correction stage continues with manual cleaning of the OCR recognition results, including deleting incorrect content, invalid content, duplicate content, spaces, and correcting incorrect segmentation and text.

[0005] However, existing document digitization technology still has many shortcomings. For example, to ensure OCR recognition accuracy, the framing process requires considerable manual effort. Furthermore, inherent technical flaws in OCR recognition often lead to errors in the recognition of misspelled characters, handwritten fonts, and complex typesetting, requiring extensive manual proofreading, which not only reduces processing efficiency but also hinders accuracy. Summary of the Invention

[0006] The purpose of the present invention is to provide a document digital processing auxiliary method and device that improves accuracy and efficiency in order to solve the above problems. The method reduces manual workload and improves processing accuracy by adding automated processing auxiliary methods to the digital processing process dominated by manual labor.

[0007] An auxiliary method for document digital processing for improving accuracy and efficiency, the method comprising: Preprocess the annotation results of the image data annotated by the user; The image data includes a number of picture words; Perform OCR recognition on the pre-processed image data, and obtain the corresponding recognition word for each image word; Determine the probability of misrecognition of each word based on the pixel distribution of the recognized word and the picture word; In response to a determination result of possible misrecognition, the corresponding picture word is intercepted from the image data to obtain a target screenshot word; Searching a target pre-stored screenshot word from a text library based on the target screenshot word; In response to finding the target pre-stored screenshot word, replacing the corresponding recognized word with the word corresponding to the target pre-stored screenshot word; In response to not finding the pre-stored screenshot word, marking the corresponding recognition word.

[0008] In the above-mentioned document digitization processing auxiliary method for improving accuracy and efficiency, the annotation includes drawing a frame and marking a frame; The preprocessing of the annotation results of the user-annotated image data includes: The position of the pull box is identified and judged, and in response to not meeting the established rules, the size of the corresponding pull box is adjusted to make it meet the established rules.

[0009] In the above-mentioned document digitization processing auxiliary method for improving accuracy and efficiency, the width of the first draw frame of each page / column is used as the standard draw frame; Identifying and judging the position of the pull box includes judging whether the horizontal position and width of the subsequent pull box are consistent with the standard pull box. If so, it is considered to comply with the established rules. Otherwise, the corresponding pull box is adjusted to make its horizontal position and width consistent with the standard pull box.

[0010] In the above-mentioned document digitization processing auxiliary method for improving accuracy and efficiency, the frame position recognition and judgment includes the judgment of the frame line position and the judgment of the spacing between adjacent pictures in the frame; The established rules include that no image characters are cut off by the frame line and the distance between adjacent image characters in the frame is less than a set distance threshold; Adjusting the corresponding pull-box size includes adjusting the pull-box size so that the truncated picture characters are circled within the pull-box or excluded from the pull-box; adjusting the pull-box size so that the picture characters that are relatively close to the pull-box edge and have a spacing greater than a set spacing threshold with adjacent picture characters are excluded from the pull-box.

[0011] In the above-mentioned document digitization auxiliary method for improving accuracy and efficiency, preprocessing the annotation results of the user-annotated image data also includes: Search the first and last coordinates of the text in each pull-down box in turn, and determine whether the text in the corresponding pull-down box is an independent paragraph based on the first and / or last coordinates; In response to the non-independent paragraph, the same paragraph as the previous pull-box is added to the next pull-box.

[0012] In the above-mentioned document digitization processing auxiliary method for improving accuracy and efficiency, the method for determining whether the drawn box is an independent paragraph includes: Determine the coordinate positions of the rightmost character and the leftmost character according to the overall pixel distribution of each page / column of the image data, and accordingly determine a standard end-of-text horizontal coordinate and a standard beginning-of-text horizontal coordinate for each page / column; Find the first coordinate and the last coordinate of each text in the pull box in turn. The first coordinate is the coordinate of the first word in the pull box, and the last coordinate is the coordinate of the last word in the pull box. If the horizontal coordinate of the end of the current pull-box meets the corresponding standard horizontal coordinate position of the end of the text, and the horizontal coordinate of the beginning of the text of the next pull-box also meets the standard horizontal coordinate position of the beginning of the text, the next pull-box is considered to be a non-independent paragraph.

[0013] In the above-mentioned document digitization processing auxiliary method for improving accuracy and efficiency, the text library is pre-stored with a number of error correction text pairs, each of which includes a pre-stored screenshot word and its corresponding text; Searching for a target pre-stored screenshot word from a text library based on the target screenshot word includes: Comparing the pixel distribution of the target screenshot word with several pre-stored screenshot words; Extracting the pre-stored screenshot word with the highest pixel distribution similarity and above a degree threshold as the target pre-stored screenshot word; The method further comprises: In response to not finding the pre-stored screenshot word, marking the corresponding recognized word and saving its corresponding target screenshot word; In response to the user's storage operation based on the target screenshot word, the target screenshot word and the text provided by the user are stored in the text library as a new error correction text pair.

[0014] In the above-mentioned document digitization auxiliary method for improving accuracy and efficiency, the first OCR recognition is performed on the pre-processed image data based on the first OCR module; The method further comprises: The text after the first OCR recognition and error correction process is handed over to the user for manual verification; Using the text that has been proofread by the user as the first text document; Performing a second OCR recognition on the pre-processed image data based on the second OCR module to generate a second text document; Comparing the recognized words in the first text document and the second text document one by one, and based on the comparison result, highlighting the recognized words in the first text document that are inconsistent with the recognition result of the second text document; The first OCR module and the second OCR module are different OCR modules, and preferably the recognition accuracy of the second OCR module is higher than that of the first OCR module.

[0015] A document digitization auxiliary device for improving accuracy and efficiency, comprising a marking module, a first OCR module, and a verification module, as well as a marking result processing module located between the marking module and the first OCR module, and an automatic error correction module located between the first OCR module and the verification module; The annotation result processing module is used to pre-process the annotation results of the image data annotated by the user; the image data includes a plurality of picture words; The automatic error correction module includes an error recognition module and an error correction module; an error recognition module, configured to determine the possibility of error recognition of each word based on the pixel distribution of the recognition word and the picture word; the recognition word is obtained by performing OCR recognition on the pre-processed image data by the first OCR module; The error correction module is used to extract the corresponding picture word of the recognition word that may be incorrectly recognized from the image data to obtain the target screenshot word; search for the target pre-stored screenshot word from the text library based on the target screenshot word; in response to finding the target pre-stored screenshot word, replace the corresponding recognition word with the text corresponding to the target pre-stored screenshot word; in response to not finding the pre-stored screenshot word, mark the corresponding recognition word.

[0016] A document digitization processing auxiliary device for improving accuracy and efficiency, further comprising a second OCR module and a verification result processing module; A second recognition module is used to perform a second OCR recognition on the pre-processed image data to generate a second text document; The verification result processing module is used to take the text verified by the user as the first text document, compare the recognition words of the first text document and the second text document one by one, and based on the comparison result, highlight the recognition words in the first text document that are inconsistent with the recognition results of the second text document.

[0017] The advantages of the present invention are: A pre-processing step is added between the annotation module and the first OCR module to assist manual work, making the manual framing process relatively casual and reducing the burden of manual annotation. At the same time, it helps generate relatively standardized annotation results, reducing errors caused by human operation, thereby improving the annotation accuracy and even the accuracy of subsequent OCR recognition, and reducing the additional costs caused by errors and rework. Since high marking accuracy can be guaranteed without strict marking by the user, the auxiliary processing of this link can improve the marking speed while ensuring the marking accuracy; A self-correction mechanism is designed during the OCR-to-text process. This mechanism uses pixel distribution to determine possible misrecognition results and uses the captured text to find more accurate replacement text. This can effectively improve the accuracy of the OCR-to-text process and significantly reduce misrecognized words, thereby alleviating the proofreading burden of subsequent manual proofreading and improving proofreading efficiency while also increasing proofreading accuracy. A second OCR recognition process is introduced to utilize the different recognition characteristics of different OCR modules to perform a complementary and comprehensive check by comparing the two texts, thereby further ensuring the accuracy of the document digitization results; Through auxiliary processing design in multiple links, the labor cost is effectively reduced and the project quality is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a method flow chart of the document digitization processing auxiliary method for improving accuracy and efficiency in the first embodiment of the present invention; Figure 2 Schematic diagram of the frame size adjustment process in the annotation result preprocessing process in the first embodiment of the present invention; Figure 3 Schematic diagram of the paragraph appending process in the annotation result preprocessing process in the first embodiment of the present invention; Figure 4 Schematic diagram of a Chinese character library according to a first embodiment of the present invention; Figure 5 Schematic diagram of the automatic error correction process of the OCR text conversion process in the first embodiment of the present invention; Figure 6 1 is a schematic diagram of an automatic error correction process when no replaceable text is found during the OCR text conversion process in the first embodiment of the present invention; Figure 7 This is a result block diagram of the document digitization processing auxiliary device for improving accuracy and efficiency in the second embodiment of the present invention; Figure 8 This is a schematic diagram of processing a non-standard frame-drawing situation and pre-processing the marking results in the third embodiment of the present invention; Figure 9 This is a schematic diagram of processing in the third embodiment of the present invention in the face of another non-standard frame-drawing situation and pre-processing of the marking results for this situation. DETAILED DESCRIPTION

[0019] Example 1 like Figure 1 As shown, the present invention provides an auxiliary method for document digitization processing that improves accuracy and efficiency. It is still mainly manual, but through a series of auxiliary processing means, it greatly reduces the manual processing workload and improves processing accuracy. The specific method is as follows: 1. Preprocess the annotation results of the image data annotated by the user.

[0020] The image data is obtained based on the scanning and basic processing of the physical document, and the image data contains a number of picture words. Scanning the physical document and basic processing to obtain image data is the first step in the current process of digitalizing the physical document. This solution does not propose improvements to this, so it will not be described here. The annotation process is the process of dividing the various parts of the document into columns, frames, and marking frames. The purpose is to divide the image into areas according to different text contents to define content information, such as title frame marking, paragraph segment frame marking, etc. Generally, the entire text content needs to be framed and marked, which is done manually and is consistent with the existing technology, so it will not be described here.

[0021] 2. Use the first OCR module to perform the first OCR recognition on the pre-processed image data. Each word in the image is recognized and the corresponding recognized word is obtained. The goal of OCR recognition is to mark the content of the box in the process and mark it as the same paragraph. The recognized text content will also be classified into the same paragraph.

[0022] 3. Determine the possibility of misrecognition of each word based on the pixel distribution of the recognized word and the picture word. Specifically, the possibility of misrecognition of each word is determined based on the similarity of pixel distribution. In this embodiment, when the similarity is preferably not less than 92%, the recognition is considered correct. On the contrary, when it is less than 92%, the recognition is considered incorrect, and the judgment result at this time is "possible misrecognition". When put into use, the similarity percentage used to divide the recognition into correct or incorrect is determined according to the specific circumstances, such as the characteristics of the OCR module of the digital processing tool used, the specific calculation method of the pixel distribution similarity, etc.

[0023] 4. In response to a possible misrecognition determination result, extract the corresponding image word from the image data to obtain a target screenshot word; and search a target pre-stored screenshot word from a character library based on the target screenshot word; 5. In response to finding the target pre-stored screenshot word, replacing the corresponding recognized word with the text corresponding to the target pre-stored screenshot word; In response to not finding the pre-stored screenshot word, marking the corresponding recognition word.

[0024] In step 1, the preprocessing of the user-annotated image data includes: The position of the pull box is identified and judged, and in response to the failure to meet the established rules, the size of the corresponding pull box is adjusted to make it meet the established rules. Figure 2 As shown: The width of the first pull-box on each page / column is used as the standard pull-box. Figure 2 In the text, "The development and improvement of computer network technology..." is the first pull-down box, which is used as the standard pull-down box.

[0025] Identifying and judging the position of the pull box includes judging whether the horizontal position and width of the subsequent pull box are consistent with the standard pull box. If so, it is considered to conform to the established rules; otherwise, adjust the corresponding pull box to make its horizontal position and width consistent with the standard pull box. Figure 2 The pull box where the content "With the advent of the Internet era..." is located is the second pull box. See Figure 2 On the left side, after comparison, the horizontal position and width are inconsistent with the standard pull box. Therefore, this method adjusts its horizontal position and size according to the standard pull box to obtain Figure 2 The result on the right side. Such adjustments are made for the entire page / column, and finally a more standardized pull box result can be obtained. The user only needs to ensure that the first box is standard. After testing, after this optimization process, the recognition accuracy of the words in the edge pictures can be significantly improved. The effect of significantly improving the annotation accuracy without significantly increasing the user's workload is achieved.

[0026] Furthermore, the preprocessing of the annotation results for the image data annotated by the user also includes: According to the overall pixel distribution of each page / column of the image data, determine a standard end horizontal coordinate and a standard start horizontal coordinate for each page / column. The horizontal coordinate of the rightmost word in each page / column is used as the standard end horizontal coordinate, and the horizontal coordinate of the leftmost word is used as the standard start horizontal coordinate. For example Figure 3 In it, the horizontal coordinates where "全" and "系" are located in the left column are determined as the standard end horizontal coordinates of the left column, and the horizontal coordinates where "不" and "木" are located in the right column are determined as the standard start horizontal coordinates of the right column.

[0027] Successively find the start coordinate and end coordinate of the text in each pull box. Here, the start coordinate refers to the coordinate of the first word in the pull box, and the end coordinate refers to the coordinate of the last word in the pull box.

[0028] When the end horizontal coordinate of the previous pull box conforms to the corresponding standard end horizontal coordinate position, and the start horizontal coordinate of the subsequent pull box also conforms to the standard start horizontal coordinate, it is considered that the subsequent pull box is a non-independent paragraph, and the subsequent pull box is appended as the same paragraph as the previous pull box. This situation generally occurs when pagination and column separation are performed. Therefore, when put into use, this judgment can be made for the first pull box and the last pull box of each page / column. It should be noted that conforming to the corresponding standard end horizontal coordinate position and standard start horizontal coordinate means being within the set error range from the standard end horizontal coordinate position and standard start horizontal coordinate, rather than being exactly the same.

[0029] Since the steps of preprocessing the annotation results are performed after the user's annotation, whether there is column separation can be directly known based on the user's pull box, and the implementation is relatively simple, so it will not be elaborated here.

[0030] With traditional tools, when you draw a box around the text, each time you draw a box, the tool will automatically mark the paragraphs. For example, when you draw the first box, it will mark paragraph 1, and when you draw the second box, it will mark paragraph 2. Sometimes, the text content of two adjacent boxes is the same paragraph. During the manual box drawing and marking process, there will be an option to append a paragraph, and the user can append the latter box as a continuation of the paragraph of the previous box. However, in actual marking, it is easy to draw a box and forget to append it. The above design of this solution targets this situation by automatically appending paragraphs based on format recognition, which can effectively solve the problem of forgetting to append during manual marking, reduce incorrect marking, and improve OCR recognition effect.

[0031] like Figure 3 As shown in the figure, a document page is displayed in two columns. The left side is the content in the left column, and the right side is part of the content in the right column. The content in the right frame is the same paragraph as the left frame. The user forgot to add it during the frame-drawing process, causing the system tool to automatically mark it as "Paragraph 2". Figure 3 The result above. After judging that the contents of the two boxes are the same paragraph, we pre-process the result of adding paragraphs to them, and finally get Figure 3 Results below.

[0032] Further, if Figure 4 As shown, the text library contains several error correction text pairs, each of which includes a pre-stored screenshot word and its corresponding text. Step 3 specifically includes: like Figure 5 As shown, the image data in the picture word It was identified as "find", and after comparing the pixel distribution, it was considered that it might be misidentified. Extract the target screenshot from the image data . Compare target screenshot words The pixel distribution of several pre-stored screenshot words in the text library is finally found from the text library. Screenshot of satisfaction and goals The condition that the pixel distribution similarity is the highest and higher than the degree threshold, the pre-stored screenshot fonts found will be As the target pre-stored screenshot word, then use the target pre-stored screenshot word The corresponding text "I" replaces the recognition result "find".

[0033] When no pre-stored screenshot word is found, the corresponding recognition word is marked and its corresponding target screenshot word is saved, and an association relationship between the recognition word and the target screenshot word is established so that the user can subsequently store it in the database. Figure 6 As shown, there is currently no text library that matches The similarity of the pixel distribution of the pre-stored screenshot characters is higher than the degree threshold, so the recognition result "valley" is highlighted.

[0034] When the user receives the text after OCR recognition and error correction processing, the user will perform manual verification and cleaning on the text content, including deleting spaces, correcting errors, etc. In this process, under the improvement of this solution, the user can correct and store the marked recognition characters. For example, when selecting "valley" for storage, the system will retrieve the corresponding target screenshot character and then use the text "rong" input by the user as the target screenshot character as the corresponding text, and the two are stored in the character library as a new pair of error correction text pairs. The next time OCR faces and when it misrecognizes it, it can automatically correct and correct it.

[0035] Furthermore, this method also includes: Taking the text processed by the user's verification as the first text document; Based on the second OCR module, perform a second OCR recognition on the preprocessed image data to generate a second text document.

[0036] Compare the recognition characters of the first text document and the second text document one by one. Based on the comparison result, highlight the recognition characters in the first text document that are inconsistent with the recognition results of the second text document.

[0037] The first OCR module and the second OCR module are different OCR modules. Preferably, the recognition accuracy of the second OCR module is higher than that of the first OCR module. The first OCR module is a module自带 by the digital processing tool system, and the second OCR module is an OCR module added in this solution to improve the processing accuracy.

[0038] After the first step of ORC recognition and automatic error correction, it is checked and adjusted manually. After the final manual processing and inspection, the final processing result will be obtained. In the traditional case, this is the final result. In this embodiment, after this step, a final full inspection process is added, and cross-comparison is performed using the different recognition characteristics of different OCR modules to further improve the accuracy of the result.

[0039] Embodiment 2 This embodiment provides a literature digital processing auxiliary device for improving accuracy and efficiency, which is used to implement the method described in Embodiment 1. As Figure 7As shown, it includes a labeling module 1, a first OCR module 2, and a verification module 3. In particular, a labeling result processing module 4 is designed between the labeling module 1 and the first OCR module 2, an automatic error correction module 5 is designed between the first OCR module 2 and the verification module 3, a verification result processing module 7 is designed after the verification module 3, and a second OCR module 6 is also provided after the labeling result processing module 4.

[0040] Among them, the annotation module 1, the first OCR module 2, and the verification module 3 are modules inherent in the digital processing tool. The unique feature of this solution is that it proposes the annotation result processing module 4, the automatic error correction module 5, the second OCR module 6, and the verification result processing module 7 on this basis, which can significantly improve the efficiency and accuracy of users using digital processing tools to digitize documents. It is important to note that the focus of this solution is on the automated auxiliary processing steps in several links. Whether to add human inspection and verification to each link and the number of human quality inspections are not within the scope of this solution.

[0041] The annotation result processing module 4 is used to pre-process the annotation result of the image data annotated by the user, where the image data contains a number of picture words; The automatic error correction module 5 includes an error recognition module 51 and an error correction module 52; The error recognition module 51 is used to determine the possibility of error recognition of each word based on the pixel distribution of the recognition word and the picture word; the recognition word is obtained by the first OCR module 2 performing OCR recognition on the pre-processed image data; The error correction module 52 is used to extract the corresponding picture word of the recognition word that may be incorrectly recognized from the image data to obtain the target screenshot word; search for the target pre-stored screenshot word from the text library based on the target screenshot word; in response to finding the target pre-stored screenshot word, replace the corresponding recognition word with the text corresponding to the target pre-stored screenshot word; in response to not finding the pre-stored screenshot word, mark the corresponding recognition word.

[0042] A second recognition module is used to perform a second OCR recognition on the image data pre-processed by the annotation result processing module 4 to generate a second text document; The verification result processing module 7 is used to treat the text verified by the user as the first text document, compare the recognition words of the first text document and the second text document one by one, and based on the comparison result, highlight the recognition words in the first text document that are inconsistent with the recognition results of the second text document.

[0043] Through the above semi-automated improvements, auxiliary processing and pre-processing modules are added in multiple links to realize automatic adjustment, correction and verification of annotations and texts, effectively improving the accuracy of annotation results and the accuracy of text conversion results, thereby improving the overall level of data quality, while reducing the annotation burden of staff and improving the efficiency of digital work processing.

[0044] Example 3 This example is similar to Example 1, except that in this example, as Figure 8 and Figure 9 shown, the preprocessing of the annotation result for adjusting the bounding box size of the image data annotated by the user includes: Identifying and judging the position of the bounding box, including judging the position of the bounding box line and the spacing between adjacent picture words within the bounding box; Judging whether it conforms to the established rules. The established rules include that no picture word is truncated by the bounding box line and the spacing between adjacent picture words within the box is less than the set spacing threshold. When there is any situation of "picture word being truncated by the bounding box line" or "spacing between adjacent picture words within the box being greater than or equal to the set spacing threshold", it is considered not to conform to the established rules.

[0045] In response to not conforming to the established rules, adjust the corresponding bounding box size. First, adjust the bounding box size so that no picture word is truncated by the bounding box, including adjusting the bounding box size so that the truncated picture word is enclosed within the bounding box or excluded outside the bounding box. As Figure 8 shown, "of computer network technology" is truncated by the bounding box line. After adjustment, there is no truncation situation.

[0046] Then adjust according to the word spacing, adjust the bounding box size so that the picture word with a spacing greater than the set spacing threshold from the adjacent picture word is excluded outside the bounding box. As Figure 9 shown, on the left is one column of content, and on the right is another column of content. The spacing between "dai" and "mu" in the figure is significantly larger than the spacing between adjacent words in a normal paragraph, and "mu" is relatively closer to the bounding box line than "dai", so it is excluded outside the bounding box. After adjustment, the content on the right part is excluded outside the bounding box. If there is no situation where the picture is truncated by the bounding box line, directly enter the second-step adjustment.

[0047] Specifically, for the situation where the bounding box line truncates the picture word, it can be set to default to enclose it within the box or exclude it outside the box. In this example, it is preferably determined whether to pull the picture word into the box or exclude it from the box according to the pixel distribution ratio of the picture word inside and outside the box. Specifically, if the pixel proportion inside the box is more than that outside the box, adjust the bounding box size so that the relevant picture word is enclosed within the box, otherwise exclude it outside the box.

[0048] This example has higher flexibility compared to Example 1. It can not only adjust the horizontal size but also the vertical size. When the user is drawing the bounding box, they can be more casual not only in the horizontal direction but also in the vertical direction. And there is no strict requirement for the accuracy of the user's bounding box drawing, and the user does not need to spend energy to ensure the standardization of the first box.

[0049] The specific embodiments described herein are merely illustrative of the spirit of the present invention. Persons skilled in the art may make various modifications, additions, or substitutions to the described specific embodiments without departing from the spirit of the present invention or exceeding the scope of the appended claims.

[0050] Although this document frequently uses terms such as annotation module 1, first OCR module 2, verification module 3, annotation result processing module 4, automatic error correction module 5, error identification module 51, error correction module 52, second OCR module 6, and verification result processing module 7, the use of other terms is not excluded. These terms are used solely to more conveniently describe and explain the essence of the present invention; interpreting them as any additional limitations would be contrary to the spirit of the present invention.

Claims

1. A document digitization processing auxiliary method for improving accuracy and efficiency, characterized in that: The method includes: Preprocess the annotation results of the image data annotated by the user; The image data includes a number of picture words; Perform OCR recognition on the pre-processed image data, and obtain the corresponding recognition word for each image word; Determine the probability of misrecognition of each word based on the pixel distribution of the recognized word and the picture word; In response to a determination result of possible misrecognition, the corresponding picture word is intercepted from the image data to obtain a target screenshot word; Searching a target pre-stored screenshot word from a text library based on the target screenshot word; In response to finding the target pre-stored screenshot word, replacing the corresponding recognized word with the word corresponding to the target pre-stored screenshot word; In response to not finding the pre-stored screenshot word, marking the corresponding recognition word.

2. The document digitization processing auxiliary method for improving accuracy and efficiency according to claim 1 is characterized in that: The marking includes a draw frame and a mark frame; The preprocessing of the annotation results of the user-annotated image data includes: The position of the pull box is identified and judged, and in response to not meeting the established rules, the size of the corresponding pull box is adjusted to make it meet the established rules.

3. The document digitization processing auxiliary method for improving accuracy and efficiency according to claim 2 is characterized in that: The width of the first drawer on each page / column is used as the standard drawer; Identifying and judging the position of the pull box includes judging whether the horizontal position and width of the subsequent pull box are consistent with the standard pull box. If so, it is considered to comply with the established rules. Otherwise, the corresponding pull box is adjusted to make its horizontal position and width consistent with the standard pull box.

4. The document digitization processing auxiliary method for improving accuracy and efficiency according to claim 2 is characterized in that: The frame position recognition and judgment includes the judgment of the frame line position and the judgment of the spacing between adjacent pictures in the frame; The established rules include that no image characters are cut off by the frame line and the distance between adjacent image characters in the frame is less than a set distance threshold; Adjusting the size of the corresponding draw frame includes adjusting the size of the draw frame so that the truncated image words are circled within the draw frame or excluded from the draw frame; The size of the pull frame is adjusted so that the picture characters that are relatively close to the edge of the pull frame and have a distance from adjacent picture characters greater than a set distance threshold are excluded from the pull frame.

5. The document digitization processing auxiliary method for improving accuracy and efficiency according to claim 3 or 4, characterized in that: The preprocessing of the annotation results of the user-annotated image data also includes: Search the first and last coordinates of the text in each pull-down box in turn, and determine whether the text in the corresponding pull-down box is an independent paragraph based on the first and / or last coordinates; In response to the non-independent paragraph, the same paragraph as the previous pull-box is added to the next pull-box.

6. The document digitization processing auxiliary method for improving accuracy and efficiency according to claim 5 is characterized in that: Methods for determining whether a pull-box is an independent paragraph include: Determine the coordinate positions of the rightmost character and the leftmost character according to the overall pixel distribution of each page / column of the image data, and accordingly determine a standard end-of-text horizontal coordinate and a standard beginning-of-text horizontal coordinate for each page / column; Find the first coordinate and the last coordinate of each text in the pull box in turn. The first coordinate is the coordinate of the first word in the pull box, and the last coordinate is the coordinate of the last word in the pull box. If the horizontal coordinate of the end of the current pull-box meets the corresponding standard horizontal coordinate position of the end of the text, and the horizontal coordinate of the beginning of the text of the next pull-box also meets the standard horizontal coordinate position of the beginning of the text, the next pull-box is considered to be a non-independent paragraph.

7. The document digitization processing auxiliary method for improving accuracy and efficiency according to claim 1 is characterized in that: The text library contains a plurality of error correction text pairs, each of which includes a pre-stored screenshot word and its corresponding text; Searching for a target pre-stored screenshot word from a text library based on the target screenshot word includes: Comparing the pixel distribution of the target screenshot word with several pre-stored screenshot words; Extracting the pre-stored screenshot word with the highest pixel distribution similarity and above a degree threshold as the target pre-stored screenshot word; The method further comprises: In response to not finding the pre-stored screenshot word, marking the corresponding recognized word and saving its corresponding target screenshot word; In response to the user's storage operation based on the target screenshot word, the target screenshot word and the text provided by the user are stored in the text library as a new error correction text pair.

8. The document digitization processing auxiliary method for improving accuracy and efficiency according to claim 1 is characterized in that: performing the first OCR recognition on the preprocessed image data based on the first OCR module; The method further comprises: The text after the first OCR recognition and error correction process is handed over to the user for manual verification; Using the text that has been proofread by the user as the first text document; Performing a second OCR recognition on the pre-processed image data based on the second OCR module to generate a second text document; Comparing the recognized words in the first text document and the second text document one by one, and based on the comparison result, highlighting the recognized words in the first text document that are inconsistent with the recognition result of the second text document; The first OCR module and the second OCR module are different OCR modules.

9. A document digitization processing auxiliary device for improving accuracy and efficiency, comprising a marking module (1), a first OCR module (2) and a verification module (3), characterized in that: It also includes a marking result processing module (4) located between the marking module (1) and the first OCR module (2), and an automatic error correction module (5) located between the first OCR module (2) and the verification module (3); The annotation result processing module (4) is used to pre-process the annotation result of the image data annotated by the user; the image data includes a plurality of picture words; The automatic error correction module (5) includes an error recognition module (51) and an error correction module (52); an error recognition module (51) for judging the possibility of error recognition of each word based on the pixel distribution of the recognition word and the picture word; the recognition word is obtained by performing OCR recognition on the pre-processed image data by the first OCR module (2); An error correction module (52) is used to intercept the corresponding picture word of the recognition word that may be incorrectly recognized from the image data to obtain the target screenshot word; Searching a target pre-stored screenshot word from a text library based on the target screenshot word; In response to finding the target pre-stored screenshot word, replacing the corresponding recognized word with the word corresponding to the target pre-stored screenshot word; In response to not finding the pre-stored screenshot word, marking the corresponding recognition word.

10. The document digitization processing auxiliary device for improving accuracy and efficiency according to claim 9, characterized in that: It also includes a second OCR module (6) and a verification result processing module (7); A second recognition module is used to perform a second OCR recognition on the pre-processed image data to generate a second text document; The verification result processing module (7) is used to use the text verified by the user as the first text document, compare the recognition words of the first text document and the second text document one by one, and based on the comparison result, highlight the recognition words in the first text document that are inconsistent with the recognition results of the second text document.

Images