A method, apparatus, and system for automatically creating nested tables from unstructured cell associations for table-based documents.

Abstract

To provide a method, apparatus and system for automatic association between keywords and content by finding relations between them in a given form.SOLUTION: A method is as follows. In some forms containing keywords and content, there may be nested levels of keywords, also referred to as a hierarchy. Content in the forms may be associated with one or more keywords in one or more of the nested levels, or in the hierarchy. Identifying keywords in adjacent cells in a table (with a nested keyword being either to the right of or below another keyword) enables distinguishing between keywords and content in filled forms, and enables correct association of content with keywords.SELECTED DRAWING: Figure 3

Claims

1. In response to form input, a deep learning model is used to identify multiple lines from one or more features of the form and to extract those lines. Using the extracted lines, identify and extract multiple connected regions, Identifying a table from the aforementioned multiple linked regions, Extracting multiple cells from the aforementioned identified table, Using the positions and sizes of multiple adjacent cells in the identified table, the multiple cells in the identified table are grouped together. The specified table is to form multiple rows and multiple columns, Using the grouping and one of the formed rows and columns, a hierarchy is formed in the specified cells in the identified table. Using the hierarchy formed, identify multiple adjacent keywords in the identified table, Using the hierarchy formed, the contents of the identified table are identified, It includes associating each identified adjacent keyword with each identified content, The aforementioned features include a form auto-registration method that can be single-color, multi-color, or grayscale.

2. The form automatic registration method according to claim 1, wherein the formed hierarchy of the identified table is a column hierarchy, and each identified content is displayed under a plurality of identified adjacent keywords.

3. The form automatic registration method according to claim 1, wherein the formed hierarchy of the identified table is a row hierarchy, and each identified content is displayed to the right of a plurality of identified adjacent keywords.

4. The form automatic registration method according to any one of claims 1 to 3, wherein the deep learning model is selected from the group consisting of convolutional neural networks.

5. The form automatic registration method according to any one of claims 1 to 4, wherein the aforementioned feature is selected from the group consisting of single color, multiple color, and grayscale, and the identification and extraction of the multiple lines includes identifying one or more shade regions of the form as single color, multiple color, or grayscale.

6. The form automatic registration method according to claim 5, wherein the extraction of the plurality of cells includes extracting one or more cells from one or more shade regions.

7. The form automatic registration method according to claim 6, wherein, in the extraction of one or more cells from one or more shade regions, the extracted lines are further used.

8. The form automatic registration method according to any one of claims 1 to 7, wherein the formed hierarchy is represented in a format selected from the group consisting of JavaScript Object Notation (JSON), Hypertext Markup Language (HTML), and Extensible Markup Language (XML).

9. The aforementioned association means that the content is associated with a specific keyword, with a minimum cost value of C ｉｊ This includes determining by identifying C ｉｊ This can be found by the following equation The form automatic registration method according to any one of claims 1 to 8. [Math 1]

10. The aforementioned semantic distance S ｉｊ The form automatic registration method according to claim 9, wherein the distance is calculated as a distance selected from the group consisting of geometric distance, forward rank (FR), backward rank (BR), arithmetic mean of FR and BR, geometric mean of FR and BR, harmonic mean of FR and BR, Euclidean distance, word mover distance, or cosine distance.

11. One or more processors, One or more non-temporary memory devices, A form automatic registration system comprising a deep learning system that executes a deep learning model, The form automatic registration system stores one or more programs in the one or more non-temporary memory devices, and the one or more programs include instructions, and when the instructions are executed, In response to form input, a deep learning model is used to identify multiple lines from one or more features of the form and to extract the multiple lines. Using the extracted lines, identify and extract multiple connected regions, Identifying a table from the aforementioned multiple linked regions, Extracting multiple cells from the aforementioned identified table, Using the positions and sizes of multiple adjacent cells in the identified table, the multiple cells in the identified table are grouped together. The specified table is to form multiple rows and multiple columns, Using the grouping and one of the formed rows and columns, a hierarchy is formed in the specified cells in the specified table. Using the hierarchy formed, identify multiple adjacent keywords in the identified table, Using the hierarchy formed, the contents of the identified table are identified, This involves associating each identified adjacent keyword with its respective content. stomach, The aforementioned feature is an automated form registration system that can be single-color, multi-color, or grayscale.

12. The form automatic registration system according to claim 11, wherein the formed hierarchy of the identified table is a column hierarchy, and each identified content is displayed under a plurality of identified adjacent keywords.

13. The form automatic registration system according to claim 11, wherein the formed hierarchy of the identified table is a row hierarchy, and each identified content is displayed to the right of a plurality of identified adjacent keywords.

14. The form automatic registration system according to any one of claims 11 to 13, wherein the deep learning model is selected from the group consisting of convolutional neural networks.

15. The form automatic registration system according to any one of claims 11 to 14, wherein the features are selected from the group consisting of single color, multiple colors, and grayscale, and the identification and extraction of the multiple lines includes identifying one or more shade regions of the form as single color, multiple colors, or grayscale.

16. The extraction of the aforementioned cells is the extraction of one or more cells from the one or more shaded regions. The form automatic registration system according to claim 15, including the above.

17. The form automatic registration system according to claim 16, wherein, in the extraction of one or more cells from one or more shade regions, the extracted lines are further used.

18. The form automatic registration system according to any one of claims 11 to 17, wherein the formed hierarchy is represented in a format selected from the group consisting of JavaScript Object Notation (JSON), Hypertext Markup Language (HTML), and Extensible Markup Language (XML).

19. The aforementioned association means that the content is associated with a specific keyword, with a minimum cost value of C ｉｊ This includes determining by identifying C ｉｊ The form automatic registration system according to any one of claims 11 to 18, wherein the following formula is used to determine the value. [Math 2]

20. The aforementioned semantic distance S ｉｊ The form automatic registration system according to claim 19, wherein the distance is calculated as a distance selected from the group consisting of geometric distance, forward rank (FR), backward rank (BR), arithmetic mean of FR and BR, geometric mean of FR and BR, harmonic mean of FR and BR, Euclidean distance, word mover distance, or cosine distance.

Images