Generation of classification data used for classifying documents

1. A computer-implemented method for generating classification data which is used for classifying documents, the method comprising:

reading, in a memory, documents in a form of a spreadsheet;

collecting cell values in each of the documents;

finding, using a processor, in each of common or near cell locations among all or a part of the documents, one or more common cell values among the collected values;

counting, using the processor, for each of the common cell values, a number of the documents having the common cell value;

storing, if the number of the documents is equal to or larger than a predetermined number, the common cell value as a candidate header label in a memory;

calculating, using the processor, a distance between cell locations of the candidate header labels in each of the documents;

choosing, according to the calculated distance, two or more candidate header labels among the candidate header labels for each of the documents; and

storing, in a storage, one or more combinations of the chosen two or more candidate header labels (hereinafter referred to as “header”) as the classification data.

2. The method as recited in claim 1 , further comprising:

counting, for each of the header, the number of the documents having cell values corresponding to the header; and

if the number of the documents is equal to or larger than a predetermined number, replacing the classification data with the header.

3. The method as recited in claim 1 , further comprising:

calculating a similarity between or among the headers;

choosing, based on the similarity, two or more headers among the headers in each of the documents; and

replacing the classification data with a combination of the chosen two or more headers (hereinafter referred to as “header pattern”).

4. The method as recited in claim 3 , wherein the choice of the two or more candidate header labels is carried out if the similarity is equal to or larger than a predetermined value.

5. The method as recited in claim 3 , wherein the similarity is calculated with cosign similarity or edit distance.

6. The method as recited in claim 3 , further comprising:

finding out an overlapping relation of headers between or among the header patterns;

setting, after finding out the overlapping relation, a part having the overlapping relation is to a new header pattern (hereinafter referred to as “a major header pattern”); and

setting, after finding out the overlapping relation, a header pattern having the part to a derived header pattern for the major header part.

7. The method as recited in claim 6 , further comprising:

calculating a ratio of the number of documents in which the major header pattern is comprised and the number of documents in which the derived header pattern is comprised; and

if the ratio is equal to or larger than a predetermined value, replacing the derived header pattern with the major header pattern.

8. The method as recited in claim 1 , further comprising:

counting, for each of the headers, the number of the documents having cell values corresponding to the header;

if the number of the documents is equal to or larger than a predetermined number, choosing one or more candidate header labels;

if the number of chosen headers is plural in each of the documents, calculating similarity between or among the headers;

choosing, based on the similarity, two or more headers among the plural headers; and

replacing the classification data with a combination of the chosen two or more headers.

9. The method as recited in claim 1 , wherein the near cell locations are cell locations in a row, cell locations in a column, or cell locations in any row and column.

10. The method as recited in claim 1 , wherein the distance between cell locations is a distance between cells in a row, a distance between cells in a column, or a distance between cells in any row and column.

11. A system for generating classification data which is used for classifying documents, comprising:

a memory; and

a processor configured to:

read, in the memory, documents in a form of a spreadsheet and collecting cell values in each of the documents;

find, in each of common or near cell locations among all or a part of the documents, one or more common cell values among the collected values;

count, for each of the common cell values, the number of the documents having the common cell value;

store, if the number of the documents is equal to or larger than a predetermined number, the common cell value as a candidate header label in a memory;

calculate a distance between cell locations of the candidate header labels in each of the document;

choose, according to the calculated distance, two or more candidate header labels among the candidate header labels for each of the documents; and

store one or more combinations of the chosen two or more candidate header labels (hereinafter referred to as “header”) as the classification data in a storage.

12. The system as recited in claim 11 , the processor being further configured to:

count, for each of the headers, the number of the documents having cell values corresponding to the header; and

replace, if the number of the documents is equal to or larger than a predetermined number, the classification data with the header.

13. The system as recited in claim 11 , the processor being further configured to:

calculate a similarity between or among the headers;

choose, based on the similarity, two or more headers among the headers in each of the documents; and

replace the classification data with a combination of the chosen two or more headers.

14. The system as recited in claim 11 , the processor being further configured to:

count, for each of the headers, the number of the documents having cell values corresponding to the header;

choose, if the number of the documents is equal to or larger than a predetermined number, one or more headers;

calculate, if the number of chosen headers is plural in each of the documents, a similarity between or among the headers;

choose, based on the similarity, two or more headers among the plural headers; and

replace the classification data with a combination of the chosen two or more headers.

15. A non-transitory computer readable storage medium comprising a computer readable program for generating classification data which is used for classifying documents, wherein the computer readable program when executed on a computer causes the computer to perform the steps of:

reading, in a memory, documents in a form of a spreadsheet and collecting cell values in each of the documents;

finding, in each of common or near cell locations among all or a part of the documents, one or more common cell values among the collected values;

counting, for each of the common cell values, a number of the documents having the common cell value;

storing, if the number of the documents is equal to or larger than a predetermined number, the common cell value as a candidate header label in a memory;

calculating a distance between cell locations of the candidate header labels in each of the document;

choosing, according to the calculated distance, two or more candidate header labels among the candidate header labels for each of the documents; and

storing one or more combinations of the chosen two or more candidate header labels (hereinafter referred to as “header”) as the classification data in a storage.

16. The non-transitory computer readable storage medium as recited in claim 15 wherein the computer readable program when executed on the computer causes the computer to further perform the steps of:

counting, for each of the headers, the number of the documents having cell values corresponding to the header; and

replacing, if the number of the documents is equal to or larger than a predetermined number, the classification data with the header.

17. The non-transitory computer readable storage medium as recited in claim 15 , wherein the computer readable program when executed on the computer causes the computer to further perform the steps of:

calculating similarity between or among the headers;

choosing, based on the similarity, two or more headers among the headers in each of the documents; and

replacing the classification data with a combination of the chosen two or more headers (hereinafter referred to as “header pattern”).

18. The non-transitory computer readable storage medium as recited in claim 17 , wherein the computer readable program when executed on the computer causes the computer to further perform the steps of:

finding out an overlapping relation of headers between or among the header patterns;

setting, after finding out the overlapping relation, a part having the overlapping relation is to a new header pattern (hereinafter referred to as “a major header pattern”); and

setting, after finding out the overlapping relation, a header pattern having the part to a derived header pattern for the major header part.

19. The non-transitory computer readable storage medium as recited in claim 18 , wherein the computer readable program when executed on the computer causes the computer to further perform the steps of:

calculating a ratio of the number of documents in which the major header pattern is comprised and the number of documents in which the derived header pattern is comprised; and

replacing, if the ratio is equal to or larger than a predetermined value, the derived header pattern with the major header pattern.

20. The non-transitory computer readable storage medium as recited in claim 15 , wherein the computer readable program when executed on the computer causes the computer to further perform the steps of:

counting, for each of the headers, the number of the documents having cell values corresponding to the header;

choosing, if the number of the documents is equal to or larger than a predetermined number, one or more headers;

calculating, if the number of chosen headers is plural in each of the documents, a similarity between or among the headers;

choosing, based on the similarity, two or more headers among the plural headers; and

replacing the classification data with a combination of the chosen two or more headers.