Method for realizing stock right financing declaration image-text material automatic identification and data structuring

Abstract

The invention relates to the technical field of data processing, in particular to a method for realizing stock right financing application image-text material automatic identification and data structuring. Inputting image-text materials, and sorting the image-text materials to obtain table data; performing identification processing on the table data, extracting a table structure, and obtaining table information; the table information and the OCR information are fused and exported; through the mode, the manual auditing work is simplified, the auditing processing efficiency is improved, and meanwhile the auditing accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method for realizing automatic recognition and data structuring of graphic materials for equity financing declaration. Background Art

[0002] Currently, in the review of corporate loan application materials, the specific process is: the enterprise submits a loan application and related materials to the relevant business platform, including but not limited to Growth Loan, High Knowledge Loan, Stock and Bond Link, etc., and undergoes manual review after submission. The manual review requires careful review of multi-dimensional key information such as the company's qualifications, operating conditions, financial data, credit records, etc. The review is relatively cumbersome and the review processing efficiency is low. At the same time, the review relies on manual labor and is not accurate.

[0003] Therefore, it is very necessary to propose a method for automatically identifying and structuring the graphic materials of equity financing declaration, which can simplify the manual review work, improve the review processing efficiency, and improve the review accuracy. Summary of the invention

[0004] The purpose of the present invention is to provide a method for automatically identifying and structuring graphic materials for equity financing declaration, aiming to solve the technical problems in the prior art that manual review requires careful review of multi-dimensional key information such as enterprise qualifications, operating conditions, financial data, credit records, etc., the review is relatively cumbersome, the review processing efficiency is low, and the review relies on manual labor and has poor accuracy.

[0005] To achieve the above purpose, the present invention adopts a method for automatically identifying and structuring the graphic materials for equity financing declaration, comprising the following steps:

[0006] Input graphic materials, organize them, and obtain table data;

[0007] Identify and process table data, extract table structure, and obtain table information;

[0008] Merge table information and OCR information and export.

[0009] Among them, in the step of inputting graphic materials, arranging the graphic materials, and obtaining table data:

[0010] Enter the complete PDF image;

[0011] Use machine learning algorithms to classify areas in PDF images;

[0012] According to the classification information, the table structure information is extracted and the PDF image is converted into table data, which is an editable Excel file.

[0013] Among them, in the step of using a machine learning algorithm to classify the areas in the PDF image:

[0014] Category areas include text, table, image, title and list.

[0015] Among them, in the step of identifying and processing the table data, extracting the table structure, and obtaining the table information:

[0016] Text detection and table structure prediction are performed separately.

[0017] Among them, in the step of text detection:

[0018] Perform single-line text detection on table data to obtain coordinates;

[0019] Get the text results in the table data.

[0020] Among them, in the step of table structure prediction:

[0021] Get the four-point coordinates and table structure information of each Excel cell.

[0022] Among them, in the step of integrating the table information and the OCR information and exporting:

[0023] Combine the coordinates of the obtained single-line text;

[0024] Join text belonging to the same cell together;

[0025] Export table data in Excel format.

[0026] The present invention provides a method for realizing automatic recognition and data structuring of graphic materials for equity financing declaration, which obtains table data by inputting graphic materials and arranging the graphic materials; performs recognition processing on the table data, extracts the table structure, and obtains the table information; integrates the table information and OCR information, and exports them; simplifies manual review work, improves review processing efficiency, and improves review accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] 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 or the description of the prior art 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.

[0028] Figure 1 It is a flow chart of the steps of the method for realizing automatic recognition and data structuring of equity financing declaration graphic materials of the present invention.

[0029] Figure 2 It is a step flow chart of S100 of the present invention.

[0030] Figure 3 It is a flow chart of the steps S200 and S300 of the present invention. DETAILED DESCRIPTION

[0031] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0032] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0033] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0034] See also Figure 1 to Figure 3 The present invention provides a method for realizing automatic recognition and data structuring of graphic materials for equity financing declaration, comprising the following steps:

[0035] S100: inputting graphic materials, arranging the graphic materials, and obtaining table data;

[0036] In this implementation, the enterprise loan application materials are reviewed. The application materials are graphic materials, and the graphic materials are sorted. The specific process of sorting is as follows:

[0037] S101: Input the complete PDF image;

[0038] S102: using a machine learning algorithm to classify regions in the PDF image; the classified regions include text, tables, images, titles, and lists;

[0039] S103: Extracting table structure information according to the classification information, and converting the PDF image into table data, which is an editable Excel file.

[0040] In the above process, layout analysis is required. For a complete PDF image, this open source project can classify the text, table, picture, title and list areas in the document image. At the same time, the table recognition technology can be used to completely extract the table structure information, so that the table image becomes an editable Excel file. PP-Structure's layout analysis technology mainly performs layout analysis on documents in the form of pictures, dividing the document into five areas: text, title, table, picture and list. Referring to the engineering structure design of Layout-Parser, combined with PaddleDetection's open source efficient detection algorithm PP-YOLOv2, the mAP on the TableBank and PubLayNet datasets reached 93.6 and 96.2 respectively, and NVIDIA Tesla P40 took only 66.6ms, and can support users to customize training according to their own data.

[0041] S200: performing recognition processing on the table data, extracting the table structure, and obtaining the table information;

[0042] In this embodiment, recognition processing is performed on table data, table structure is extracted, and table information is obtained, and the process includes: text detection and table structure prediction.

[0043] The specific process of text detection is:

[0044] S201: Perform single-line text detection on table data to obtain coordinates;

[0045] S202: Obtain text results in the table data.

[0046] The specific process of table structure prediction is as follows:

[0047] S203: Obtain the four-point coordinates and table structure information of each Excel cell.

[0048] S300: Merge the table information and OCR information, and export them.

[0049] In this embodiment, the specific process of fusion is as follows:

[0050] S301: input the coordinates of the obtained single-line text;

[0051] S302: splicing texts belonging to the same cell together;

[0052] S303: Export the table data in Excel format.

[0053] In S200: performing recognition processing on table data, extracting table structure, obtaining table information and S300: integrating table information with OCR information and exporting, table recognition is required, and the image description model RARE based on the attention mechanism is used; the image description model RARE includes six parts, namely: text detection module, text recognition module, table structure prediction module, cell coordinate aggregation module, cell text aggregation module, Excel export module; among which:

[0054] The text detection module is used to detect single-line text on the table data to obtain coordinates;

[0055] The text recognition module is used to obtain text results in the table data;

[0056] The table structure prediction module is used to obtain the four-point coordinates and table structure information of each Excel cell;

[0057] The Cell coordinate aggregation module is used to input the coordinates of the obtained single-line text;

[0058] The Cell text aggregation module is used to splice together texts belonging to the same cell;

[0059] The Excel export module is used to export table data in Excel format.

[0060] The overall process can be divided into two parts, wherein the first part is a common OCR process, in which the text detection module detects single-line text on the table data to obtain coordinates; the text recognition module obtains the text results in the table data.

[0061] In the second part, the table image first passes through the table structure prediction module to obtain the four-point coordinates and table structure information of each Excel cell. Combined with the four-point coordinates of the single-line text box obtained in the first part, they are input into the Cell coordinate aggregation module, and then the text belonging to the same cell is spliced ​​together through the Cell text aggregation module. Finally, combined with the table structure information, the table data in Excel format is exported through the Excel export module.

[0062] In the present invention, the enterprise loan application materials are reviewed, and the application materials are graphic materials; first, the graphic materials are sorted to obtain the table data; then the table data is identified and processed, the table structure is extracted, and the table information is obtained; then the table information and OCR information are integrated and exported; the manual review work is simplified, the review processing efficiency is improved, and the review accuracy is improved.

[0063] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the contents disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary technical means in the art that are not disclosed in the present application.

[0064] It should be understood that the present application is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof.

Claims

1. A method for realizing automatic recognition and data structuring of graphic materials for equity financing declaration, characterized in that: The steps include: Input graphic materials, organize them, and obtain table data; Identify and process table data, extract table structure, and obtain table information; Merge table information and OCR information and export.

2. The method for realizing automatic recognition and data structuring of graphic materials for equity financing declaration as claimed in claim 1, characterized in that: In the step of inputting graphic materials, arranging them, and obtaining table data: Enter the complete PDF image; Use machine learning algorithms to classify areas in PDF images; According to the classification information, the table structure information is extracted and the PDF image is converted into table data, which is an editable Excel file.

3. The method for realizing automatic recognition and data structuring of graphic materials for equity financing declaration as claimed in claim 2, characterized in that: In the step of using machine learning algorithm to classify the areas in the PDF image: Category areas include text, table, image, title and list.

4. The method for realizing automatic recognition and data structuring of graphic materials for equity financing declaration as claimed in claim 2, characterized in that: In the steps of identifying and processing table data, extracting table structure, and obtaining table information: Text detection and table structure prediction are performed separately.

5. The method for realizing automatic recognition and data structuring of equity financing declaration graphic materials as claimed in claim 4, characterized in that: In the steps of text detection: Perform single-line text detection on table data to obtain coordinates; Get the text results in the table data.

6. The method for realizing automatic recognition and data structuring of equity financing declaration graphic materials as claimed in claim 5, characterized in that: In the steps of table structure prediction: Get the four-point coordinates and table structure information of each Excel cell.

7. The method for realizing automatic recognition and data structuring of equity financing declaration graphic materials as claimed in claim 6, characterized in that: In the step of fusing table information with OCR information and exporting: Combine the coordinates of the obtained single-line text; Join text belonging to the same cell together; Export table data in Excel format.

Images