Method and System of Pre-Analysis and Automated Classification of Documents

• US 20110188759A1
• Filed: 04/14/2011
• Published: 08/04/2011
• Est. Priority Date: 06/26/2003
• Status: Abandoned Application

• Current Assignee: ABBYY Production LLC (ABBYY Software)
• Original Assignee: ABBYY Production LLC (ABBYY Software)
• Inventors: Myakutin, Andrey, Filimonova, Irina, Zlobin, Sergey
• Application Number:
  US13/087,242
• Publication Number:
  US 20110188759A1
• Time in Patent Office: Days
• US Class Current: 382/195
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06F 18/24765 Rule-based classification
• CPC Class Codes: G06V 10/40 Extraction of image or video features
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/416 Extracting the logical structure, e.g. chapters, sections or page numbers; Identifying elements of the document, e.g. authors

Abstract

Automatic classification of different types of documents is disclosed. An image of a form or document is captured. The document is assigned to one or more type definitions by identifying one or more objects within the image of the document. A matching model is selected via identification of the document image. In the case of multiple identifications, a profound analysis of the document type is performed—either automatically or manually. An automatic classifier may be trained with document samples of each of a plurality of document classes or document types where the types are known in advance or a system of classes may be formed automatically without a priori information about types of samples. An automatic classifier determines possible features and calculates a range of feature values and possible other feature parameters for each type or class of document. A decision tree, based on rules specified by a user, may be used for classifying documents. Processing, such as optical character recognition (OCR), may be used in the classification process.

Claims

1. A method for a computer system to perform an analysis of document type, the method comprising:

providing to the computer system a document image;

detecting at least one feature in the document image;

assigning a text to the at least one feature in the document image;

matching the document image to one or more nodes of at least one decision tree based at least in part upon the text assigned to the at least one feature in the document image; and

associating the document image with one or more document types based at least in part upon the matching the document image to the one or more nodes of the at least one decision tree.

2. The method of claim 1 wherein the at least one decision tree is created at least partially on the basis of one or more features previously identified in a training process, wherein the training process comprises use of document samples of known document types.

3. The method of claim 2 wherein the training process further comprises:

detecting one or more features in at least one of the training document samples;

forming the at least one decision tree based at least in part upon the detected one or more features in the at least one training document samples, wherein forming the at least one decision tree comprises creating a node on the basis of the detected one or more features in the at least one training document samples; and

saving training data from the one or more of the training document samples in one or more binary formats and storing the training data for use by the computer system or another machine.

4. The method of claim 3 wherein the detecting one or more features in at least one of the training document samples includes calculating a range of values associated with each of the one or more detected features of the training document samples.

5. The method of claim 4 wherein the creating the decision tree is also based in part upon the range of values associated with each of the one or more detected features of the training documents.

6. The method of claim 1 wherein the one or more decision trees are created on the basis of rules using one or more flexible descriptions derived at least in part from the document image.

7. The method of claim 1 wherein the assigning the document image based in part upon the one or more decision trees includes associating the document image to the one or more nodes of the decision tree.

8. The method of claim 1 wherein the method further comprises:

further processing the document image after assigning it to one or more document types.

9. The method of claim 5 wherein the method further comprises further processing of the document image in accordance with its type (class) or according to a combination of types (classes) to which the document image was assigned.

10. The method of claim 1 wherein the method is performed prior to recognizing the document.

11. One or more computer readable media configured to bear a device detectable implementation of a method, the method comprising:

identifying one or more document features in a document image;

correlating one or more of the one or more document features with one or more document classes;

forming a decision tree based at least in part upon the identified one or more document features in the document, wherein forming the decision tree includes creating a node corresponding to each of the one or more document classes; and

associating with one or more of the document classes the document image based in part upon the decision tree and the document image.

12. The one or more computer readable media of claim 12, wherein the identified one or more document features are document features that were previously determined to be one or more of the most reliable document features capable of distinguishing documents, and wherein the one or more most reliable document features were previously identified by analysis of a plurality of training documents each having at least one feature different from at least one of the other training documents.

13. The one or more computer readable media of claim 12, wherein each document feature is associated with a feature type, wherein the identifying one or more document features in the plurality of training documents includes identifying a feature type for each document feature, and wherein a decision tree is formed for each of the feature types identified.

14. The one or more computer readable media of claim 13, wherein creating a node corresponding to each of the document types in each of the decision trees formed for each of the feature types identified.

15. The one or more computer readable media of claim 14, wherein a feature type is selected from a list comprising:

raster, title, image object, text string, word, unique mark, unique character, numeric code, non-human readable marking, and other.

16. The one or more computer readable media of claim 12, wherein the document features in the plurality of training documents are predefined, and wherein the identifying one or more document features in the document image includes performing optical character recognition on each of the document features in the document image.

17. The one or more computer readable media of claim 12, wherein the document image is associated with one of the one or more of the document classes based in part upon a value determined from the document image and in part upon a reliability index determined for the decision tree, wherein the reliability index is determined at least in part from the one or more document features of the plurality of training documents.

18. The one or more computer readable media of claim 12, wherein the method further comprises:

prior to the identifying the one or more document features in the document image, identifying a document object in the document image, wherein the identifying the one or more document features in the document image is identifying the one or more document features in the document object.

19. A system for classifying an unclassified document, the system comprising:

a decision tree trainer that is configured to receive a plurality of training documents, identify one or more features in the training documents, identify one or more document classes based on the one or more features in the training documents, and create a node or sub-node in the decision tree for each of the one or more document classes; and

a document classifier that is configured to classify an unclassified document based in part on one or more features identified in an image associated with the unclassified document and in part on one or more nodes of the decision tree, in part on one or more sub-nodes of the decision tree, or in part on a combination of one or more nodes of the decision tree and one or more sub-nodes of the decision tree.

20. The system of claim 19 wherein the document classifier is configured to perform a complex analysis of all or one or more portions of the image associated with the unclassified document when the document classifier classifies the unclassified document in two or more classes based upon one or more features identified the an image associated with the unclassified document.