KR102309562B1 - Apparatus for pdf table reconstruction and method thereof - Google Patents

Abstract

The present invention relates to a device for reconstructing a table included in a PDF document and an operation method thereof. In accordance with one aspect of the present invention, provided is a method for operating a PDF table reconstruction device. The method comprises the steps of: receiving a PDF document including a table to be reconstructed; extracting PDF data including coordinate information of characters and lines included in the PDF document; retrieving a table area to be reconstructed based on the PDF data; generating a primary data structure of a reconstruction table composed of a plurality of cells based on coordinate information of characters and lines corresponding to the table area to be reconstructed in the PDF data; retrieving a merged cell among a plurality of cells constituting the primary data structure; and splitting the merged cell to create a final data structure of the reconstruction table. In accordance with the present invention, by means of the PDF table construction device, a table included in a PDF document can be reconstructed.

Description

PDF table reorganization device and its operation method

The present invention relates to an apparatus for reconstructing a table included in a PDF document and a method for operating the same.

Securities companies, etc. regularly publish reports summarizing the management and financial status of companies, in which data on items such as profit and loss calculation, financial status, cash flow, and key indicators for each company can be clearly identified in a table format. It is often included in Portable Document Format (PDF) documents. However, if the outline of multiple cells constituting the table included in the PDF document is not clearly expressed or if there are cells in the form of multiple cells being merged, there is a limit to the use of other types of documents in the table included in the PDF document. will have

Therefore, it is necessary to clarify the outline of the cells and separate the merged cells in the table included in the PDF document.

It is an object of the present invention to provide a technique for reconstructing a table included in a PDF document.

Another object of the present invention is to provide a technique for reconstructing a table included in an image file of a PDF document.

According to an aspect of the present invention, there is provided a method comprising: receiving a PDF document including a table to be reconstructed; extracting PDF data including coordinate information of characters and lines included in the PDF document; retrieving a table area to be reorganized based on the PDF data; generating a primary data structure of a reorganization table composed of a plurality of cells based on coordinate information of characters and lines corresponding to a table area to be reconstructed among PDF data; searching for a merged cell among a plurality of cells constituting the primary data structure; and dividing the merge cell to generate a final data structure of the reorganization table.

In an embodiment, extracting the PDF data may include extracting the PDF data by parsing the PDF document.

In an embodiment, extracting the PDF data may include extracting the PDF data by parsing the PDF document using a PDF miner.

In one embodiment, the final data structure may be in JavaScript Object Notation (JSON) format.

In an embodiment, the merged cell is a form in which L horizontal cells are merged, and the generating of the final data structure may generate the final data structure by dividing the merged cell into L horizontal cells.

In an embodiment, the merged cell is a form in which M vertical cells are merged, and the generating of the final data structure may generate the final data structure by dividing the merged cell into M vertical cells.

In an embodiment, the merged cell is a parent cell corresponding to the N number of lower cells, and generating the final data structure may generate the final data structure by dividing the merged cell into N vertical cells.

In one embodiment, the step of searching for the table area to be reconstructed is to convert the PDF file to an image file if the table area to be reconstructed is not searched, and to include it in the image file based on a deep learning-based image analysis model for the image file The steps of extracting the coordinate information of the characters and lines extracted from the image file, searching the table area to be reconstructed based on the coordinate information of the characters and lines extracted from the image file, and generating the primary data structure include the coordinate information of the characters and lines extracted from the image file. A primary data structure of a reorganization table composed of a plurality of cells may be generated based on coordinate information of characters and lines corresponding to the table area to be reconstructed.

In an embodiment, the deep learning-based image analysis model may be MM detection.

In addition, according to another aspect of the present invention, the method comprising: receiving an image file of a PDF document; extracting coordinate information of characters and lines included in an image file based on a deep learning-based image analysis model; searching for a table area to be reconstructed based on coordinate information of characters and lines included in the image file; generating a primary data structure of a reconfiguration table composed of a plurality of cells based on coordinate information of characters and lines corresponding to a table region to be reconstructed among the coordinate information of characters and lines included in the image file; searching for a merged cell among a plurality of cells constituting the primary data structure; and generating a final data structure of the reorganization table by dividing the merge cell.

In addition, according to another aspect of the present invention, a memory for storing instructions for reconstructing a table included in a PDF document; and a processor for reconstructing a table of the PDF document according to instructions stored in the memory, wherein the instructions extract PDF data including coordinate information of characters and lines included in the PDF document, and select a table area to be reconstructed based on the PDF data. Search, create a primary data structure of a reorganization table consisting of a plurality of cells based on the coordinate information of characters and lines corresponding to the table area to be reorganized among PDF data, and merge cells among a plurality of cells constituting the primary data structure A PDF table reorganization apparatus is provided, which is a command for generating a final data structure of a reorganization table by searching for , splitting a merge cell.

In addition, according to another aspect of the present invention, a memory for storing instructions for reconstructing a table included in the image file of the PDF document; and a processor for reconstructing a table included in an image file of a PDF document according to a command stored in the memory, wherein the command extracts coordinate information of characters and lines included in the image file based on a deep learning-based image analysis model, Searches the table area to be reconstructed based on the coordinate information of the text and line included in the image file, and selects a plurality of cells based on the coordinate information of the text and line corresponding to the table area to be reconstructed among the coordinate information of the text and line included in the image file. PDF table reorganization, which is a command that creates the primary data structure of the configured reorganization table, searches for merged cells among a plurality of cells constituting the primary data structure, and splits the merged cells to generate the final data structure of the reorganized table A device is provided.

According to an embodiment of the present invention, it becomes possible to reconstruct a table included in a PDF document.

Further, according to another embodiment of the present invention, it becomes possible to reconstruct a table included in an image file of a PDF document.

1 is a flowchart of an operating method of a PDF table reorganization apparatus according to an embodiment of the present invention;
Figure 2 is a flowchart of an operating method of a PDF table reorganization apparatus according to another embodiment of the present invention.
3 to 7 are diagrams illustrating examples of a merge cell according to an embodiment of the present invention.
8 is a block diagram of a PDF table reorganization apparatus according to an embodiment of the present invention.
9 is a diagram illustrating a result of parsing PDF data according to an embodiment of the present invention;

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and will be described in detail through the detailed description. However, this is not intended to limit the present invention to specific embodiments, and it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention. In describing the present invention, if it is determined that a detailed description of a related known technology may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. Also, as used herein and in the claims, the terms "a" and "a" and "a" are to be construed to mean "one or more" in general, unless stated otherwise.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. do it with

1 is a flowchart of an operating method of a PDF table reconfiguration apparatus for reconstructing a table of a PDF document according to an embodiment of the present invention.

In step S1010, a PDF document is received. Specifically, the PDF table reconstructing apparatus may receive a PDF document including a table to be reconstructed from an external device through a wired/wireless communication method.

In step S1020, PDF data is extracted. Specifically, the PDF table reconstructing apparatus may extract PDF data including coordinate information such as characters and lines included in the PDF document from the received PDF document.

In an embodiment, the PDF table reconstructing apparatus may extract PDF data from the received PDF document by parsing the received PDF document. For example, the PDF table reconstructing apparatus may extract PDF data by parsing the PDF document using a PDF miner or the like. Data extracted by parsing using PDF miner includes arrangement information of characters and lines in the PDF document (LTRect means line information in the PDF document, LTTextLineHorizontal indicates the text coordinates and contents in the PDF document) as shown in FIG. 9 can do.

In step S1030, a table area to be reorganized is searched for. Specifically, the PDF table reorganization apparatus may search for a table area to be reconstructed based on PDF data extracted from the received PDF document.

In an embodiment, when the PDF data includes information for identifying the table area, the PDF table reorganization apparatus may check the table area to be reconstructed by using the information.

In an embodiment, the apparatus for reconstructing a PDF table may identify a table area to be reconstructed based on an arrangement interval and a position of characters and lines included in the PDF data. For example, when the vertical and horizontal arrangement of characters correspond to vertical and horizontal cells constituting the table, the PDF table reorganization apparatus may identify the table area to be reorganized. That is, when (A) of FIG. 3 is a table area to be reconstructed included in a PDF document, the PDF table reconfiguration apparatus includes horizontal items (horizontal items 1 to 3), vertical items (vertical items 1 to 3) and these By recognizing that the horizontal and vertical intervals at which the corresponding values A to G are arranged are in the form of a cell as in FIG. can be checked

In an embodiment, when the PDF data includes coordinate information of an outline of a table, the PDF table reconstructing apparatus may identify it as a table area to be reconstructed.

In step S1040, the PDF document is converted into an image file. Specifically, the PDF table reconstructing apparatus may convert the PDF document into an image file (eg, jpg) if the table area to be reconstructed is not retrieved from the PDF data.

In an embodiment, the apparatus for reconstructing a PDF table may not recognize a table area when an outline of a table included in a PDF file does not exist.

In step S1050, image analysis is performed on the image file. Specifically, the PDF table reconstructing apparatus may analyze the image file converted from the PDF file, and extract characters, lines, and coordinate information thereof included in the image file.

In an embodiment, the PDF table reconstruction apparatus may extract characters, lines, and coordinate information thereof using a deep learning-based image analysis model. For example, the PDF table reconstruction device uses a deep learning-based image analysis model such as MM detection (object detection model library of Pytorch, a deep learning framework), and uses text, lines and their coordinate information included in an image file. can be extracted.

In step S1060, a table area to be reconstructed is searched for based on the image analysis result. Specifically, the PDF table reconstructing apparatus may identify a table area to be reconstructed based on characters, lines, and coordinate information thereof extracted from an image file.

In an embodiment, the apparatus for reconstructing a PDF table may identify a table area to be reconstructed based on an arrangement interval and location of characters and lines included in characters and lines extracted from an image file and their coordinate information. For example, when the vertical and horizontal arrangement of characters correspond to vertical and horizontal cells constituting the table, the PDF table reorganization apparatus may identify the table area to be reorganized. That is, when (A) of FIG. 3 is a table area to be reconstructed included in a PDF document, the PDF table reconfiguration apparatus includes horizontal items (horizontal items 1 to 3), vertical items (vertical items 1 to 3) and these By recognizing that the horizontal and vertical intervals at which the corresponding values A to G are arranged are in the form of a cell as in FIG. can be checked

In step S1070, a primary data structure of the reorganization table is created. Specifically, the apparatus for reconstructing a PDF table may generate a primary data structure composed of a plurality of cells based on coordinate information of characters and lines corresponding to the searched table area among PDF data.

In an embodiment, the apparatus for reconstructing a PDF table may generate a primary data structure composed of a plurality of cells by allocating a size and number of cells corresponding to an arrangement interval of characters and lines corresponding to the searched table area. For example, an interval at which the letters of FIG. 3A are arranged in the horizontal and vertical directions correspond to a plurality of cells having the same size and number of cells of FIG. 3B . Accordingly, the PDF table reconstructing apparatus generates a primary data structure of the form shown in FIG. 3(B) from the coordinate information of the characters arranged as shown in FIG. 3(A).

In addition, the PDF table reconstructing apparatus may generate a primary data structure composed of a plurality of cells based on the coordinate information of the characters and lines corresponding to the searched reconstruction table among the characters, lines, and their coordinate information extracted from the image file.

In an embodiment, the primary data structure may be in various forms, such as JavaScript Object Notation (JSON), text (txt), Pandas (Python Data Analysis Library), NumPy (NPY) type data, and the like. In this case, the primary data structure may be stored according to a cell position basis, a vertical cell name basis, a horizontal cell name basis, and the like.

In step S1080, a merged cell is searched for. Specifically, the PDF table reorganization apparatus may search for a cell in which two or more cells are merged among a plurality of cells constituting the primary data structure of the reorganization table.

In an embodiment, the merged cell may be a cell in which L (a natural number greater than or equal to 2) horizontal cells are merged. For example, as shown in (A) of FIG. 4 , three horizontal cells corresponding to horizontal item 2 may be merged cells.

In an embodiment, the merged cell may be a cell in which M (a natural number greater than or equal to 2) vertical cells are merged. For example, as shown in FIG. 5A , two vertical cells corresponding to vertical item 3 may be merged cells.

In an embodiment, the merged cell may be a cell in which L horizontal cells and M vertical cells are merged. For example, as shown in FIG. 6A , four cells corresponding to vertical items 1 and 2 and horizontal items 1 and 2 may be merged cells.

In an embodiment, the merged cell may be a cell corresponding to N (a natural number greater than or equal to 2) sub-cells. For example, it may be an upper cell corresponding to two lower cells (lower cell 1 and lower cell 2) as shown in (A) of FIG. 7 .

In step S1090, the merge cell is split to create a final data structure. Specifically, the PDF table reorganization apparatus may generate the final data structure of the reorganization table by dividing the merged cell among a plurality of cells constituting the primary data structure.

In an embodiment, when the merged cell is a form in which L horizontal cells are merged, the PDF table reconstructing apparatus may generate a final data structure by dividing the merged cell into L horizontal cells. For example, when three horizontal cells are merged as shown in FIG. 4(A), the PDF table reconstructing apparatus may divide the merged cell into three horizontal cells as shown in FIG. 4(B). Also, since the values of vertical items 1 to 3 corresponding to horizontal item 2 are all the same as A, the PDF table reconstructing apparatus may assign the same value A to all divided regions.

In an embodiment, when the merged cells are merged into M vertical cells, the PDF table reconstructing apparatus may generate a final data structure by dividing the merged cells into M vertical cells. For example, when two vertical cells are merged as shown in FIG. 5A , the PDF table reconstructing apparatus may divide the PDF table into two vertical cells as shown in FIG. 5B . Also, since values of horizontal items 1 to 2 corresponding to vertical item 3 are all the same as A, the PDF table reconstructing apparatus may assign the same value A to all divided regions.

In an embodiment, when the merged cell is a merged form of K (a natural number greater than or equal to 2) horizontal cells and vertical cells, the PDF table reconstructing apparatus may generate a final data structure by dividing the merged cells into K horizontal cells and vertical cells. . For example, when the PDF table reconstructing apparatus has a form in which four horizontal cells and vertical cells are merged as shown in FIG. 6(A), it may be divided into four horizontal cells and vertical cells as shown in FIG. 6(B). . In addition, since the values of the vertical items 1 and 2 corresponding to the horizontal items 1 and 2 are all the same as A, the PDF table reconstructing apparatus may allocate the same value A to the divided regions.

In an embodiment, the PDF table reconstructing apparatus may generate a final data structure by dividing the merged cell into N cells when the merged cell is an upper cell corresponding to N lower cells. For example, the PDF table reconstructing apparatus may divide a merge cell corresponding to two sub-cells into two cells as shown in FIG. 7A . In this case, the letter (A) corresponding to the upper cell and the letter (B, C) corresponding to the lower cell may be combined (AB, AC, or BA, CA). That is, the lower cell further limits the concept of the upper cell, so for example, when A is a male, B is a teenager, and C is 20 Teenager male)", "Men in their 20s (or 20s male)" can be assigned.

2 is a flowchart of a method of operating a PDF table reconstructing apparatus for reconstructing a table of a PDF document according to another embodiment of the present invention.

In step S2010, an image file of the PDF document is received. Specifically, the PDF table reconstructing apparatus may receive an image file of a PDF document including a table to be reconstructed from an external device through a wired/wireless communication method.

In step S2020, image analysis is performed on the image file. Specifically, the PDF table reconstructing apparatus may analyze the image file converted from the PDF file, and extract characters, lines, and coordinate information thereof included in the image file.

In an embodiment, the PDF table reconstruction apparatus may extract characters, lines, and coordinate information thereof using a deep learning-based image analysis model. For example, the PDF table reconstruction apparatus may extract characters, lines, and coordinate information thereof included in an image file by using a deep learning-based image analysis model such as MM detection.

In step S2030, a table area to be reconstructed is searched for based on the image analysis result. Specifically, the PDF table reconstructing apparatus may identify a table area to be reconstructed based on characters, lines, and coordinate information thereof extracted from an image file.

In an embodiment, the apparatus for reconstructing a PDF table may identify a table area to be reconstructed based on an arrangement interval and location of characters and lines included in characters and lines extracted from an image file and their coordinate information. For example, when the vertical and horizontal arrangement of characters correspond to vertical and horizontal cells constituting the table, the PDF table reorganization apparatus may identify the table area to be reorganized. That is, when (A) of FIG. 3 is a table area to be reconstructed included in a PDF document, the PDF table reconfiguration apparatus includes horizontal items (horizontal items 1 to 3), vertical items (vertical items 1 to 3) and these By recognizing that the horizontal and vertical intervals at which the corresponding values A to G are arranged are in the form of a cell as in FIG. can be checked

In step S2040, a primary data structure of the reorganization table is created. Specifically, the apparatus for reconstructing a PDF table may generate a primary data structure composed of a plurality of cells based on coordinate information of characters and lines corresponding to the searched table area among PDF data.

In an embodiment, the apparatus for reconstructing a PDF table may generate a primary data structure composed of a plurality of cells by allocating a size and number of cells corresponding to an arrangement interval of characters and lines corresponding to the searched table area. For example, an interval at which the letters of FIG. 3A are arranged in the horizontal and vertical directions correspond to a plurality of cells having the same size and number of cells of FIG. 3B . Accordingly, the PDF table reconstructing apparatus generates a primary data structure of the form shown in FIG. 3(B) from the coordinate information of the characters arranged as shown in FIG. 3(A).

In addition, the PDF table reconstructing apparatus may generate a primary data structure composed of a plurality of cells based on the coordinate information of the characters and lines corresponding to the searched reconstruction table among the characters, lines, and their coordinate information extracted from the image file.

In one embodiment, the primary data structure may be JSON (JavaScript Object Notation) type data.

In step S2050, a merged cell is searched for. Specifically, the PDF table reorganization apparatus may search for a cell in which two or more cells are merged among a plurality of cells constituting the primary data structure of the reorganization table.

In an embodiment, the merged cell may be a cell in which L horizontal cells are merged. For example, as shown in (A) of FIG. 4 , three horizontal cells corresponding to horizontal item 2 may be merged cells.

In an embodiment, the merged cell may be a cell in which M vertical cells are merged. For example, as shown in FIG. 5A , two vertical cells corresponding to vertical item 3 may be merged cells.

In an embodiment, the merged cell may be a cell in which L horizontal cells and M vertical cells are merged. For example, as shown in FIG. 6A , four cells corresponding to vertical items 1 and 2 and horizontal items 1 and 2 may be merged cells.

In one embodiment, the merge cell may be a cell corresponding to N sub-cells. For example, it may be an upper cell corresponding to two lower cells (lower cell 1 and lower cell 2) as shown in (A) of FIG. 7 .

In step S2060, the merged cell is divided to create a final data structure. Specifically, the PDF table reorganization apparatus may generate the final data structure of the reorganization table by dividing the merged cell among a plurality of cells constituting the primary data structure.

In an embodiment, when the merged cell is a merged form of L horizontal cells, the PDF table reconstructing apparatus may generate a final data structure by dividing the merged cell into L horizontal cells. For example, when three horizontal cells are merged as shown in FIG. 4(A), the PDF table reconstructing apparatus may divide the merged cell into three horizontal cells as shown in FIG. 4(B). Also, since the values of vertical items 1 to 3 corresponding to horizontal item 2 are all the same as A, the PDF table reconstructing apparatus may assign the same value A to all divided regions.

In an embodiment, when the merged cells are merged into M vertical cells, the PDF table reconstructing apparatus may generate a final data structure by dividing the merged cells into M vertical cells. For example, when two vertical cells are merged as shown in FIG. 5A , the PDF table reconstructing apparatus may divide the PDF table into two vertical cells as shown in FIG. 5B . Also, since values of horizontal items 1 to 2 corresponding to vertical item 3 are all the same as A, the PDF table reconstructing apparatus may assign the same value A to all divided regions.

In an embodiment, when the merged cell is a merged form of K horizontal cells and vertical cells, the PDF table reconstruction apparatus may generate a final data structure by dividing the merged cells into K horizontal cells and vertical cells. For example, when the PDF table reconstructing apparatus has a form in which four horizontal cells and vertical cells are merged as shown in FIG. 6(A), it may be divided into four horizontal cells and vertical cells as shown in FIG. 6(B). . In addition, since the values of the vertical items 1 and 2 corresponding to the horizontal items 1 and 2 are all the same as A, the PDF table reconstructing apparatus may allocate the same value A to the divided regions.

In an embodiment, the PDF table reconstructing apparatus may generate a final data structure by dividing the merged cell into N cells when the merged cell is an upper cell corresponding to N lower cells. For example, the PDF table reconstructing apparatus may divide a merge cell corresponding to two sub-cells into two cells as shown in FIG. 7A . In this case, the letter (A) corresponding to the upper cell and the letter (B, C) corresponding to the lower cell may be combined (AB, AC, or BA, CA). That is, the lower cell further limits the concept of the upper cell, so for example, when A is a male, B is a teenager, and C is 20 Teenager male)", "Men in their 20s (or 20s male)" can be assigned.

8 is a block diagram of a PDF table reconfiguration apparatus 800 according to an embodiment of the present invention.

As shown in FIG. 8 , the PDF table reconstruction apparatus 800 includes at least one or more of a processor 810 , a memory 820 , a storage 830 , a user interface input unit 840 , and a user interface output unit 850 . It may include elements, which may communicate with each other via bus 860 . In addition, the PDF table reconstruction apparatus 800 may also include a network interface 870 for connecting to a network. The processor 810 may be a CPU or a semiconductor device that executes processing instructions stored in the memory 820 and/or the storage 830 . The memory 820 and the storage 830 may include various types of volatile/nonvolatile storage media. For example, the memory may include ROM 824 and RAM 825 .

In an embodiment, the memory may include instructions for performing the method shown in FIGS. 1 and 2 , and the processor 810 may reconstruct a table included in the PDF document according to the instructions stored in the memory.

In addition, when the PDF table reconstructing device 800 needs to extract a value corresponding to a specific character from a plurality of PDF documents, the PDF table reconstructing device 800 parses the first PDF document to search for a specific character, and The position of a value corresponding to a character can be set based on a specific character. For example, the PDF table reorganization device 800 sets the position of the value corresponding to the date by saying that 2019.07.01 (value) corresponding to the date (specific item) in the PDF document is located at the lower left corner based on the date. can Thereafter, when a PDF document from which a value corresponding to a specific item is to be extracted is received, the PDF table reconstructing device 800 searches the location of the specific item by parsing the PDF file, and extracts the data present at the set location from the searched location. can For example, the PDF table reconstructing apparatus 800 may parse the received PDF document to search for a date, and extract data located at the lower left corner based on the searched date. In this way, the PDF table reconstruction apparatus 800 makes it possible to quickly extract a value corresponding to a specific item from a plurality of PDF documents.

So far, the present invention has been looked at focusing on the embodiments thereof. Those of ordinary skill in the art to which the present invention pertains will understand that the present invention can be implemented in a modified form without departing from the essential characteristics of the present invention. Therefore, the disclosed embodiments are to be considered in an illustrative rather than a restrictive sense. The scope of the present invention is indicated in the claims rather than the foregoing description, and all differences within the scope equivalent thereto should be construed as being included in the present invention.

Claims (13)

A method of operating an apparatus for reconstructing a table of a PDF document, the method comprising:
receiving a PDF document comprising a table to be reconstructed;
extracting PDF data including coordinate information of characters and lines included in the PDF document;
searching the table area to be reorganized based on the PDF data;
generating a primary data structure of a reorganization table including a plurality of cells based on coordinate information of characters and lines corresponding to the table area to be reconstructed among the PDF data;
searching for a merged cell among the plurality of cells constituting the primary data structure; and
dividing the merged cell to generate a final data structure of the reorganization table;
including,
The merged cell has one value,
Creating the final data structure comprises:
Creating the final data structure by assigning the one value to each cell into which the merged cell is divided,
If the table area to be reorganized is not found in the step of searching for the table area to be reorganized,
Converts the PDF document to an image file, extracts coordinate information of characters and lines included in the image file based on a deep learning-based image analysis model for the converted image file, and coordinates information of the extracted characters and lines Perform the steps of creating the primary data structure by checking the table area to be reorganized based on the spacing and position of the characters and lines included in the
Creating the primary data structure comprises:
Among the coordinate information of characters and lines extracted from the image file, the arrangement interval and position of the line corresponding to the table area to be reconstructed, and the horizontal and vertical intervals at which the vertical and horizontal arrangement of characters and values based thereon are arranged are divided into vertical cells and horizontal cells. A method of recognizing and creating a primary data structure of a reorganized table composed of multiple cells.
According to claim 1,
The step of extracting the PDF data comprises extracting the PDF data by parsing the PDF document.
According to claim 1,
The extracting of the PDF data comprises extracting the PDF data by parsing the PDF document using a PDF miner.
According to claim 1,
The final data structure is in the format of at least one of JSON (JavaScript Object Notation), text (txt), Pandas (Python Data Analysis Library), and NumPy (NPY).
According to claim 1,
The merged cell is a form in which L horizontal cells are merged,
wherein generating the final data structure creates the final data structure by dividing the merged cell into L horizontal cells.
According to claim 1,
The merged cell is a merged form of M vertical cells,
wherein generating the final data structure creates the final data structure by dividing the merged cell into M vertical cells.
According to claim 1,
The merged cell is an upper cell corresponding to N number of lower cells,
wherein generating the final data structure creates the final data structure by dividing the merged cell into N vertical cells.
delete According to claim 1,
The deep learning-based image analysis model is MM detection, the method.
A method of operating an apparatus for reconstructing a table of a PDF document, the method comprising:
receiving an image file of the PDF document;
extracting coordinate information of characters and lines included in the image file based on a deep learning-based image analysis model;
retrieving a table area to be reconstructed based on an arrangement interval and position of characters and lines included in the coordinate information of characters and lines extracted from the image file;
A reorganization table consisting of a plurality of cells by recognizing the arrangement interval and position of the line corresponding to the searched table area to be reconstructed, the vertical and horizontal arrangement of characters and the horizontal interval and the vertical interval at which the values are arranged as vertical cells and horizontal cells creating a primary data structure;
searching for a merged cell among the plurality of cells constituting the primary data structure; and
dividing the merged cell to generate a final data structure of the reorganization table;
including,
The merged cell has one value,
Creating the final data structure comprises:
and generating the final data structure by respectively assigning the one value to each cell into which the merged cell is divided.
11. The method of claim 10,
The deep learning-based image analysis model is MM detection, the method.
a memory for storing instructions for reconstructing tables included in the PDF document; and
A processor for reconstructing the table of the PDF document according to the instructions stored in the memory,
The command is
Extracting PDF data including coordinate information of characters and lines included in the PDF document, searching for the table area to be reconstructed based on the PDF data, and selecting text and lines corresponding to the table area to be reconstructed among the PDF data Create a primary data structure of a reorganization table consisting of a plurality of cells based on the coordinate information,
If the table area to be reorganized is not found,
Converts the PDF document to an image file, extracts coordinate information of characters and lines included in the image file based on a deep learning-based image analysis model for the converted image file, and coordinates information of the extracted characters and lines Create the primary data structure by checking the table area to be reorganized based on the spacing and position of the text and lines included in the
The primary data structure is
Among the coordinate information of characters and lines extracted from the image file, the arrangement interval and position of the line corresponding to the table area to be reconstructed, and the horizontal and vertical intervals at which the vertical and horizontal arrangement of characters and values based thereon are arranged are divided into vertical cells and horizontal cells. Recognized and composed of a plurality of cells,
A command for generating a final data structure of the reorganization table by searching for a merged cell among the plurality of cells constituting the primary data structure, dividing the merged cell,
PDF table reorganization device.
a memory for storing instructions for reconstructing a table included in an image file of a PDF document; and
A processor for reconstructing a table included in the image file of the PDF document according to the instructions stored in the memory,
The command is
Coordinate information of characters and lines included in the image file is extracted based on a deep learning-based image analysis model, and reconfigured based on the spacing and location of characters and lines included in the coordinate information of characters and lines extracted from the image file Searches the table area to be reconstructed, and recognizes the arrangement spacing and position of lines corresponding to the searched table area to be reconstructed, the vertical and horizontal arrangement of characters and the horizontal spacing and the vertical spacing at which values are arranged as vertical cells and horizontal cells, and plural generating a primary data structure of a reorganization table composed of cells of command,
The merged cell has one value,
and generating the final data structure by assigning the one value to each cell into which the merged cell is divided.

Images