Back to results
US 12,567,275 B1Granted
Computing system and method for extracting unstructured document data
AssigneeVelocity EHS Holdings, Inc.
InventorsJulia Penfield, Aatish Suman
Life of this patent
Estimating the term…
Abstract
A computing system and method are configured to receive an input in a portable document format, generate a raster image of the input, and detect a form type of the input based on the raster image. In response to detecting that the raster image of the input indicates a first form type, the computing system and method are configured to partition the raster image into sections corresponding to document regions containing document entities targeted for retrieval for a specific form type, generate sub-images based on bounding coordinates of the sections, and apply selected document data identification and retrieval computational techniques to extract the document entities from the sub-images.
Claims18
A system deployed within a communication network, the system comprising:
- a computing device deployed within the communication network, the computing device comprising:
- a non-transitory computer-readable storage medium configured to store an application program,
- a processor coupled to the non-transitory computer-readable storage medium and configured to control a plurality of modules to execute instructions of the application program to obtain an input in a portable document format (PDF) via a first user interface element of the application program;
- and
- a computing server system configured to:
- receive the input from the computing device via a first application programming interface (API) call,
- determine document entities to be identified and retrieved from the input in accordance with each of a plurality of form types,
- analyze each of the document entities to determine a plurality of document data identification and retrieval computational techniques,
- generate a raster image of the input in accordance with a selected dot per inch (DPI),
- extract, from the raster image, textual content represented by a list of tokens and coordinates of each token,
- encode at least pixel-level information of the raster image, the textual content and the coordinates of each token into another input to a machine learning/deep learning classification model via a second API call,
- detect a form type of the input based on the raster image of the input by determining the form type of the input based at least in part on probabilities generated by the machine learning/deep learning classification model, the probabilities indicating a likelihood that the input corresponds to each of the plurality of form types,
- in response to detecting that the raster image of the input indicates one of the plurality of form types, partition the raster image into a plurality of sections corresponding to document regions containing the document entities targeted for retrieval for the one of the plurality of form types, generate sub-images based on bounding coordinates of the plurality of sections, and apply one or more of the plurality of document data identification and retrieval computational techniques to extract the document entities from the sub-images, and
- save processing results of the plurality of document data identification and retrieval computational techniques in identifying and retrieval of each of the document entities in at least one database.
11 more elements
A computer-implemented method, comprising:
- obtaining, by a processor of a computing device deployed within a communication network, an input in a portable document format (PDF) via a user interface element of an application program downloaded and stored on a non-transitory computer-readable storage medium of the computing device;
- receiving, in a computing server system deployed within the communication network, the input from the computing device via a first application programming interface (API) call;
- determining, by the computing server system, document entities to be identified and retrieved from the input in accordance with each of a plurality of form types;
- analyzing, by the computing server system, each of the document entities to determine a plurality of document data identification and retrieval computational techniques;
- generating, by the computing server system, a raster image of the input in accordance with a selected dot per inch (DPI);
- extracting, by the computing server system, textual content represented by a list of tokens and coordinates of each token from the raster image;
- encoding, by the computing server system, at least pixel-level information of the raster image, the textual content, and the coordinates of each token into another input to a machine learning/deep learning classification model via a second API call;
- detecting, by the computing server system, a form type of the input based on the raster image of the input by determining the form type of the input based at least in part on probabilities generated by the machine learning/deep learning classification model, the probabilities indicating a likelihood that the input corresponds to each of the plurality of form types;
- in response to detecting that the raster image of the input indicates one of the plurality of form types, partitioning, by the computing server system, the raster image into a plurality of sections corresponding to document regions containing the document entities targeted for retrieval for the one of the plurality of form types, generating sub-images based on bounding coordinates of the plurality of sections, and applying one or more of the plurality of document data identification and retrieval computational techniques to extract the document entities from the sub-images;
- and
- saving, by the computing server system, processing results of the plurality of document data identification and retrieval computational techniques in identifying and retrieval of each of the document entities in at least one database.
8 more elements
Dependencies are inferred from the claim text.