Back to results
US 2022/0067320 A1Published application
Document classification neural network and ocr-to-barcode conversion
AssigneeKYOCERA Document Solutions Inc.
InventorOleg Y. Zakharov
Life of this patent
Estimating the term…
Abstract
Document classification techniques are disclosed that convert text content extracted from documents into graphical images and apply image classification techniques to the images. A graphical image of the text (such as a bar-code) may be generated and applied to improve the performance of document classification, bypassing NLP and utilizing more efficient localized OCR than in conventional approaches.
Claims20
A system comprising:
- a document parser configured to extract a content sub-region from a document;
- an image generator configured to transform the content sub-region into a graphical code;
- a neural network trained to output one or more classifiers in response to receiving the graphical code;
- and
- a search engine operable on the one or more classifiers to return one or more document identifiers in response to a search request.
2 more elements
A method comprising:
- transforming a sub-region of a document from content in a rasterized format to content in a text format using optical character recognition, wherein the optical character recognition is localized to the sub-region;
- bypassing natural language processing to transform the content in the text format into a graphical code;
- and
- training a neural network with the graphical code labeled with one or more classifiers.
1 more element
A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a computer, cause the computer to:
- transform a sub-region of a document from content in a rasterized format to content in a text format and topological features, wherein the transform is localized to the sub-region;
- bypass natural language processing to transform the content in the text format and the topological features into a graphical code;
- and
- embed the graphical code in the document.
1 more element
Dependencies are inferred from the claim text.