System and method for separation and classification of unstructured documents

1. A system, comprising:

a processor;

a non-transitory computer-readable medium configured with a page-level recognition model trained to recognize a document start page and a document end page of an unstructured document and a document-level recognition model trained to recognize a document type and stored instructions executable by the processor to:

for each unstructured document page in a set of unstructured document pages, determine a page-level classification for the unstructured document page by processing content extracted from the unstructured document page using the page-level recognition model to determine the page-level classification;

based on the page-level classifications for the unstructured document pages in the set of unstructured document pages, identify a sequence of unstructured document pages from a first unstructured document page classified as the document start page to a second unstructured document page classified as the document end page;

generate an indication of the sequence of unstructured document pages;

combine the sequence of unstructured document pages into a separated unstructured document based on the indication;

store the separated unstructured document as an unclassified document;

determine a first document-level classification for the separated unstructured document created from the sequence of unstructured document pages by processing content extracted from the separated unstructured document using the document-level recognition model; and

automatically update metadata for the separated unstructured document with the first document-level classification.

2. The system of claim 1 , wherein the stored instructions are further executable to:

based on the indication, copy the sequence of unstructured document pages to a document folder corresponding to the separated unstructured document, the document folder accessible by a capture system to merge the sequence of unstructured document pages to create the separated unstructured document.

3. The system of claim 1 , wherein the stored instructions are further executable to:

separate a merged document containing structured and unstructured documents into a first set of classified documents and the set of unstructured document pages; and

individually store each unstructured document page in the set of unstructured document pages.

4. The system of claim 3 , wherein each unstructured document page in the set of unstructured document pages comprises a page image from the merged document.

5. The system of claim 3 , wherein the stored instructions are further executable to store the first set of classified documents and the separated unstructured document in a hierarchy established for the merged document.

6. The system of claim 1 , wherein the stored instructions are further executable to identify an orphan page from the set of unstructured document pages based on the page-level classifications for the unstructured document pages in the set of unstructured document pages.

7. The system of claim 6 , wherein the stored instructions are further executable to determine a second document-level classification for the orphan page by processing content extracted from the orphan page using the document-level recognition model.

8. A computer program product comprising a non-transitory computer-readable medium storing instructions executable by a processor to:

for each unstructured document page in a set of unstructured document pages, determine a page-level classification for the unstructured document page by processing content extracted from the unstructured document page using a page-level recognition model to determine the page-level classification for the unstructured document page, the page-level recognition model trained to recognize a document start page and a document end page of an unstructured document;

based on the page-level classifications for the unstructured document pages in the set of unstructured document pages, identify a sequence of unstructured document pages from a first unstructured document page classified as the document start page to a second unstructured document page classified as the document end page;

generate an indication of the sequence of unstructured document pages;

combine the sequence of unstructured document pages into a separated unstructured document based on the indication;

store the separated unstructured document as an unclassified document;

determine a first document-level classification for the separated unstructured document created from the sequence of unstructured document pages by processing content extracted from the separated unstructured document using a document-level recognition model trained to recognize a document type; and

automatically update metadata for the separated unstructured document with the first document-level classification.

9. The computer program product of claim 8 , wherein the stored instructions are further executable to:

based on the indication, copy the sequence of unstructured document pages to a document folder corresponding to the separated unstructured document, the document folder accessible by a capture system to merge the sequence of unstructured document pages to create the separated unstructured document.

10. The computer program product of claim 8 , wherein the stored instructions are further executable to:

separate a merged document containing structured and unstructured documents into a first set of classified documents and the set of unstructured document pages; and

individually store each unstructured document page in the set of unstructured document pages.

11. The computer program product of claim 10 , wherein each unstructured document page in the set of unstructured document pages comprises a page image from the merged document.

12. The computer program product of claim 10 , wherein the stored instructions are further executable to store the first set of classified documents and the separated unstructured document in a hierarchy established for the merged document.

13. The computer program product of claim 8 , wherein the stored instructions are further executable to identify an orphan page from the set of unstructured document pages based on the page-level classifications for the unstructured document pages in the set of unstructured document pages.

14. The computer program product of claim 13 , wherein the stored instructions are further executable to determine a second document-level classification for the orphan page by processing content extracted from the orphan page using the document-level recognition model.

15. A method for classifying unstructured document comprising:

for each unstructured document page in a set of unstructured document pages, determining, by a processor, a page-level classification for the unstructured document page by processing content extracted from the unstructured document page using a page-level recognition model to determine the page-level classification for the unstructured document page, the page-level recognition model trained to recognize a document start page and a document end page of an unstructured document;

based on the page-level classifications for the unstructured document pages in the set of unstructured document pages, identifying, by the processor, a sequence of unstructured document pages from a first unstructured document page classified as the document start page to a second unstructured document page classified as the document end page;

combining, by the processor, the sequence of unstructured document pages into a separated unstructured document;

storing, by the processor, the separated unstructured document as an unclassified document;

determining, by the processor, a first document-level classification for the separated unstructured document created from the sequence of unstructured document pages by processing content extracted from the separated unstructured document using a document-level recognition model trained to recognize a document type; and

automatically updating, by the processor, metadata for the separated unstructured document with the first document-level classification.

16. The method of claim 15 , further comprising the processor performing:

generating an indication of the sequence of unstructured document pages; and

based on the indication, copying the sequence of unstructured document pages to a document folder corresponding to the separated unstructured document, the document folder accessible by a capture system to merge the sequence of unstructured document pages to create the separated unstructured document.

17. The method of claim 15 , further comprising the processor performing:

separating a merged document containing structured and unstructured documents into a first set of classified documents and the set of unstructured document pages; and

individually storing each unstructured document page in the set of unstructured document pages.

18. The method of claim 17 , wherein each unstructured document page in the set of unstructured document pages comprises a page image from the merged document.

19. The method of claim 17 , further comprising the processor storing the first set of classified documents and the separated unstructured document in a hierarchy established for the merged document.

20. The method of claim 15 , further comprising the processor identifying an orphan page from the set of unstructured document pages based on the page-level classifications for the unstructured document pages in the set of unstructured document pages.

21. The method of claim 20 , further comprising the processor determining a second document-level classification for the orphan page by processing content extracted from the orphan page using the document-level recognition model.