Automatic document classification using text and images

• US 7,039,856 B2
• Filed: 09/30/1998
• Issued: 05/02/2006
• Est. Priority Date: 09/30/1998
• Status: Expired due to Fees

• Current Assignee: Ricoh Company Limited
• Original Assignee: Ricoh Company Limited
• Inventors: Hull, Jonathan J., Peairs, Mark, Cullen, John F.
• Primary Examiner(s): Hong, Stephen S.
• Assistant Examiner(s): HUYNH, CONG LAC T
• Application Number:
  US09/163,848
• Publication Number:
  US 20010042085A1
• Time in Patent Office: 2,771 Days
• Field of Search: 715/500, 715/530, 715/526, 715/514, 715528-529, 706/15, 706/20, 707/526, 707/514, 707/102, 707528-529, 707/5, 707/201, 345/624, 345/629
• US Class Current: 715/200
• CPC Class Codes: G06F 16/353 into predefined classes
• CPC Class Codes: G06F 16/5846 using extracted text

Abstract

A method and apparatus for automatic document classification using text and images. The present invention provides a method and apparatus for automatic document classification based on text and image. A new document is analyzed based on textual content as well as visual appearance. The new document is automatically stored in one or more mirror directories in which the new document would most likely be stored by the user of the device if the new document were placed manually. Determination of the most likely directories is based on an analysis of multiple documents stored by the user in various directories. The mirror directories are components of a mirror directory structure, which is a copy of a pre-existing directory structure, such as the user'"'"'s hard drive. By storing the new document automatically, the user is relieved of the duty of manually selecting a directory for the new document.

Claims

1. A method for document classification comprising:

using a first directory structure mirroring a second directory structure used by a user for storing documents;

analyzing content of the documents within the second directory structure to determine a plurality of document classes within the second directory structure, the plurality of document classes indicating a user approach to placing documents in the second directory structure;

determining a document classification profile associated with the first directory structure based on the plurality of document classes;

analyzing content of a previously unclassified electronic document to determine a textual profile and a graphical profile of the electronic document;

generating a classification of the document based on the textual profile and the graphical profile; and

storing the electronic document in one or more directories within the first directory structure based on the classification of the document and the document classification profile associated with the first directory structure, to resemble the user approach to placing the documents in the second directory structure.

2. The method defined in claim 1 wherein the first directory structure comprises a hierarchy of documents mirroring in a similar fashion an organization in the second directory structure representing a pre-existing memory storing documents.

3. The method defined in claim 2 wherein the pre-existing memory comprises a hard disk.

4. The method of claim 1, wherein analyzing content of an electronic document comprises:

determining characteristic words of the document;

determining a frequency for each characteristic word; and

building a frequency table based on the frequency associated with each characteristic word.

5. The method of claim 1, wherein analyzing content of an electronic document comprises:

determining a point set corresponding to the electronic document, wherein points of the point set correspond to end points of lines;

determining a density of points within the point set;

generating the graphical profile based, at least in part, on the density of points within the point set.

6. The method of claim 1, wherein generating a classification of the document based on the textual profile and the graphical profile comprises combining results from textual and graphical analysis using a Borda Count.

7. The method defined in claim 1 further comprising building the first directory structure by building a hierarchy of documents based on organization of documents stored in a hard drive.

8. The method defined in claim 1 further comprising building the first directory structure by extracting graphical and text features from documents in a directory-based memory to obtain a document classification profile of each subdirectory in the directory-based memory.

9. A software product including a machine-readable medium having stored thereon sequences of instructions, which, when executed by a processor, cause the processor to:

use a first directory structure mirroring a second directory structure used by a user for storing documents;

analyze content of the documents within the second directory structure to determine a plurality of document classes within the second directory structure, the plurality of document classes indicating a user approach to placing documents in the second directory structure;

determine a document classification profile associated with the first directory structure based on the plurality of document classes;

analyze content of a previously unclassified electronic document to determine a textual profile and a graphical profile of the electronic document;

generate a classification of the document based on the textual profile and the graphical profile; and

store the electronic document in one or more directories within the first directory structure based on the classification of the document and the document classification profile associated with the first directory structure, to resemble the user approach to placing the documents in the second directory structure.

10. The machine-readable medium of claim 9, wherein the sequences of instructions that cause the processor to analyze content of an electronic document further comprise sequences of instructions that cause the processor to:

determine characteristic words of the document;

determine a frequency for each characteristic word; and

build a frequency table based on the frequency associated with each characteristic word.

11. The machine-readable medium of claim 9, wherein the sequences of instructions that cause the processor to analyze content of an electronic document further comprise sequences of instructions that cause the processor to:

determine a point set corresponding to the electronic document, wherein points of the point set correspond to end points of lines;

determine a density of points within the point set;

generate the graphical profile based, at least in part, on the density of points within the point set.

12. The machine-readable medium of claim 9, wherein the sequences of instructions that cause the processor to generate a classification of the document based on the textual profile and the graphical profile further comprises sequences of instructions that cause the processor to combine results from textual and graphical analysis using a Borda Count.

13. A method for document classification comprising:

analyzing content of documents within a pre-existing directory structure to determine a plurality of document classes within the pre-existing directory structure, the plurality of document classes indicating a user approach to placing documents in the pre-existing directory structure;

determining a document classification profile of the pre-existing directory structure based on the plurality of document classes;

generating a mirror directory structure based on the pre-existing document directory structure;

receiving a previously unclassified electronic document;

analyzing content of the electronic document to determine a textual profile and a graphical profile of the electronic document; and

placing the electronic document at a certain storage location in the mirror directory structure based on the document classification profile of the pre-existing document directory structure, the textual profile of the document, and the graphical profile of the document, to resemble the user approach to placing the documents in the pre-existing directory structure.

14. The method of claim 13, wherein analyzing content of documents within the pre-existing document directory structure further comprises:

recursively descending the pre-existing document directory structure;

generating a list of directories in the pre-existing document directory structure; and

examining files in the directories of the pre-existing document directory structure.

15. The method of claim 13 wherein the pre-existing document directory structure is a hard disk directory structure.

16. The method of claim 13 wherein generating a mirror directory structure based on the pre-existing document directory structure comprises generating a document directory structure having a set of directories and relationships equivalent to the pre-existing document directory structure.

17. The method of claim 13 wherein placing the electronic document in the mirror directory structure comprises:

determining a primary directory in the preexisting document directory structure in which the document is to be placed based on the document classification profile of the pre-existing document directory structure; and

storing the document in a directory of the mirror directory structure that corresponds to the primary directory in the pre-existing document directory structure.

18. The method of claim 17 further comprising:

determining a secondary directory in the pre-existing document directory in which the document is to be placed based on the document classification profile of the pre-existing document directory structure; and

storing the document in a directory of the mirror directory structure that corresponds to the secondary directory in the pre-existing document directory structure.

19. A computer-readable medium having stored thereon sequences of instructions which, when executed by a processor, cause the processor to:

analyze content of documents within a pre-existing directory structure to determine a plurality of document classes within the pre-existing directory structure, the plurality of document classes indicating a user approach to placing documents in the pre-existing directory structure;

determine a document classification profile of the preexisting directory structure based on the plurality of document classes;

generate a mirror directory structure based on the pre-existing document directory structure;

receive a previously unclassified electronic document;

analyze content of the electronic document to determine a textual profile and a graphical profile of the electronic document; and

place the electronic document at a certain storage location in the mirror directory structure based on the document classification profile of the pre-existing document directory structure, the textual profile of the document, and the graphical profile of the document, to resemble the user approach to placing the documents in the pre-existing directory structure.

20. The computer-readable medium of claim 19, wherein the sequences of instructions that cause the processor to analyze content of documents within a pre-existing document directory structure further comprise sequences of instructions that cause the processor to:

recursively descending the pre-existing document directory structure;

generating a list of directories in the pre-existing document directory structure; and

examining files in the directories of the preexisting document directory structure.

21. The computer-readable medium of claim 19, wherein the sequences of instructions that cause the processor to generate a mirror directory structure further comprise sequences of instructions that cause the processor to generate a document directory structure having a set of directories and relationships equivalent to the pre-existing document directory structure.

22. The computer-readable medium of claim 19, wherein the sequences of instructions that cause the processor to place a document in the mirror directory structure further comprise sequences of instructions that cause the processor to:

determine a primary directory in the pre-existing document directory structure in which the document is to be placed based on the document classification profile of the pre-existing document directory structure; and

store the document in a directory of the mirror directory structure that corresponds to the primary directory in the pre-existing document directory structure.

23. The computer-readable medium of claim 22 further comprising sequences of instructions that cause the processor to:

determine a secondary directory in the pre-existing document directory in which the document is to be placed based on the document classification profile of the pre-existing document directory structure; and

store the document in a directory of the mirror directory structure that corresponds to the secondary directory in the pre-existing document directory structure.

24. An apparatus comprising:

means for analyzing content of documents within a pre-existing directory structure to determine a plurality of document classes within the pre-existing directory structure, the plurality of document classes indicating a user approach to placing documents in the pre-existing directory structure;

means for determining a document classification profile of the pre-existing directory structure based on the plurality of document classes;

means for generating a mirror directory structure based on the pre-existing document directory structure;

means for receiving a previously unclassified electronic document;

means for analyzing content of the electronic document to determine a textual profile and a graphical profile of the electronic document; and

means for placing the electronic document at a certain storage location in the mirror directory structure based on the document classification profile of the pre-existing document directory structure, the textual profile of the document, and the graphical profile of the document, to resemble the user approach to placing the documents in the pre-existing directory structure.

25. The apparatus of claim 24, wherein means for analyzing content of documents within the pre-existing document directory structure further comprises:

means for recursively descending the pre-existing document directory structure;

means for generating a list of directories in the pre-existing document directory structure; and

means for examining files in directories of the pre-existing document directory structure.

26. The apparatus of claim 24, wherein means for generating a mirror directory structure comprises means for generating a document directory structure having a set of directories and relationships equivalent to the pre-existing document directory structure.

27. The apparatus of claim 24, wherein means for placing a document in the mirror directory structure comprises:

means for determining a primary directory in the pre-existing document directory structure in which the document is to be placed based on the document classification profile of the pre-existing document directory structure; and

means for storing the document in a directory of the mirror directory structure that corresponds to the primary directory in the pre-existing document directory structure.

28. The apparatus of claim 27 further comprising:

means for determining a secondary directory in the pre-existing document directory in which the document is to be placed based on the document classification profile of the pre-existing document directory structure; and

means for storing the document in a directory of the mirror directory structure that corresponds to the secondary directory in the pre-existing document directory structure.

29. A document processing system comprising:

a document scanning device;

a document storage device coupled to the document scanning device, wherein the document storage device has a pre-existing document directory structure and a mirror document directory structure organized based on the pre-existing document directory structure; and

a processor coupled to the document scanning device and to the document storage device, wherein the processor is to analyze content of documents within the pre-existing document directory structure to determine a plurality of document classes in the pre-existing document directory structure, the plurality of document classes indicating a user approach to placing documents in the pre-existing directory structure, to determine a document classification profile of the pre-existing document directory structure based on the plurality of document classes, to analyze content of a document scanned by the document scanning device, to determine which directory in the mirror document directory structure the scanned document is to be placed based on the analysis of the content of the scanned document and the document classification profile of the pre-existing document directory structure, and to store the scanned document in the determined directory in the mirror document directory structure to resemble the user approach to placing the documents in the pre-existing directory structure.

30. The document processing system of claim 29 wherein the processor is further to determine a secondary directory in the pre-existing document directory structure in which the document is to be placed and to store the document in a corresponding secondary directory in the mirror document directory structure.

31. The document processing system of claim 29 wherein the processor is to analyze content of a document scanned by the document scanning device based on image and textual content of the scanned document.