Automated document extraction and classification

1. A method comprising:

receiving a source file containing a plurality of documents which, to a computer, initially are indistinguishable from each other;

obtaining a parsed file comprising a plurality of identified sub-types of documents from the source file comprising the plurality of documents indistinguishable from each other, by:

applying, by the computer, a first classification stage to the source file using a convolutional neural network image classification to identify source documents in the plurality of documents and to produce a partially parsed file having a plurality of identified source documents, the partially parsed file comprising sub-images corresponding to the plurality of identified source documents;

applying, by the computer, a second classification stage, comprising a natural language processing artificial intelligence, to sets of text in bounding boxes of the sub-images, to classify each of the plurality of identified source documents as a corresponding sub-type of document, each of the sets of text corresponding to one of the sub-images, and to produce the parsed file having the plurality of identified sub-types of documents; and

using the parsed file for further computer processing.

2. The method of claim 1 , further comprising:

transmitting the parsed file to a computer processing service.

3. The method of claim 2 , further comprising:

processing, by the computer processing service, each corresponding sub-type of document according to types of the sets of text contained in the sub-images.

4. The method of claim 1 , further comprising:

after applying the first classification stage, but before applying the second classification stage, applying an optical character recognition algorithm to each of the sub-images to produce the plurality of sets of text.

5. The method of claim 1 , wherein using the parsed file comprises:

transmitting the parsed file to a plurality of specialized tax preparation computer services.

6. The method of claim 5 , further comprising:

extracting form fields of each of the plurality of identified sub-types of documents based on the corresponding sub-type of document.

7. The method of claim 6 , further comprising:

exporting the form fields to corresponding ones of the different specialized tax preparation computer services.

8. The method of claim 7 , further comprising:

performing computerized tax preparation algorithms specific to individual tax entities using form fields corresponding to the individual tax entities.

9. The method of claim 1 , wherein using comprises transmitting the parsed file to another algorithm executing on the computer.

10. The method of claim 1 , wherein the source file comprises one of an image and a portable document file.

11. The method of claim 1 , further comprising:

applying the bounding boxes around the sub-images.

12. The method of claim 1 , wherein using parsed file comprises performing, by the computer, at least one of storing the parsed file on a non-transitory computer readable storage medium, displaying the parsed file, and transmitting the parsed file to another computer.

13. A non-transitory computer readable storage medium storing instructions for a computer to classify a document, the instructions, when executed by a computer processor, comprising:

program code for receiving a source file containing a plurality of documents which, to a computer, initially are indistinguishable from each other;

program code for obtaining a parsed file comprising a plurality of identified sub-types of documents from the source file comprising the plurality of documents indistinguishable from each other, by:

applying a first classification stage to the source file using a vision-based convolutional neural network image classification to identify source documents in the plurality of documents, and to produce a partially parsed file having a plurality of identified source documents, the partially parsed file being composed of sub-images corresponding to the plurality of identified source documents;

program code for applying a second classification stage, comprising a natural language processing artificial intelligence, to sets of text in bounding boxes of the sub-images, to classify each of the plurality of identified source documents as a corresponding sub-type of document, each of the sets of text corresponding to one of the sub-images, and to produce the parsed file having the plurality of identified sub-types of documents; and

program code for using the parsed file for further computer processing.

14. The non-transitory computer readable storage medium of claim 13 , wherein the instructions further comprise:

program code for, after applying the first classification stage, but before applying the second classification stage, applying an optical character recognition algorithm to each of the plurality of sub-images to produce the plurality of sets of text.

15. The non-transitory computer readable storage medium of claim 13 , wherein the program code for using the parsed file comprises:

program code for transmitting the parsed file to a plurality of specialized tax preparation computer services.

16. The non-transitory computer readable storage medium of claim 13 , wherein the program code for applying the first classification stage further includes:

program code for applying the bounding boxes around the plurality of sub-images.

17. A system comprising:

a data storage device comprising a storage medium, the data storage device containing a plurality of source files, each of the plurality of source files having a corresponding plurality of source documents, each of the plurality of source documents which to a computer, initially are indistinguishable from each other; and

a computer programmed with a multi-stage classifier configured to obtain a parsed file comprising a plurality of identified sub-types of documents from one source file of the plurality of source files, the computer comprising:

a first classifier configured to identify each of the plurality of source documents as a corresponding plurality of sub-images by applying a vision-based convolutional neural network to each of the plurality of source files, each of the plurality of sub-images corresponding to one of the plurality of source documents, wherein the first classifier is thereby configured to produce a partially parsed file, and

a second classifier configured to identify a plurality sub-types of documents by applying a natural language processing artificial intelligence process to the corresponding plurality of sub-images, wherein the second classifier is thereby configured to produce the parsed file having the plurality of identified sub-types of documents having corresponding sets of text,

wherein the computer is further programmed with a computer processing service configured to further sort the sub-types of documents into different kinds of sub-types of documents, and then process the sets of text according to the different kinds of sub-types of documents.

18. The system of claim 17 , further comprising:

an optical character recognition classifier configured to identify the sets of text in the plurality of sub-images.

19. The system of claim 17 , wherein the first classifier is further configured to apply bounding boxes around each of the plurality of sub-images.

20. The system of claim 17 , wherein the plurality of source documents are tax forms, wherein sub-types of documents are specific tax forms, and wherein the kinds of sub-types of documents are groups of specific tax forms having identical sub-types.

21. The system of claim 20 , wherein the computer processing service is configured to automatically perform computerized tax preparation services on the sets of text.

22. The system of claim 17 , wherein the computer processing service is further configured to perform at least one from a group consisting of storing the parsed file on a non-transitory computer readable storage medium, displaying the parsed file, and transmitting the parsed file to another computer.