Data extraction service
AssigneeTarget Brands, Inc.
InventorsNeha Dixit, Nishtha Gupta, Vidya Hungud, Ashok J, Mohammed Jafer, Arun Patil, Suresh Saini, Debraj Singha
Life of this patent
Estimating the term…
Abstract
A data extraction service is disclosed. The data extraction service may receive an image that includes a data table. The data extraction service may perform optical character recognition on the image. The data extraction service may label text in the image to identify the table. The data extraction service may generate a digital data table corresponding to the data table in the image. The data extraction service may format the digital data table and provide the formatted digital data table to a downstream entity.
Claims20
A method for extracting data from an image, the method comprising:
- receiving the image, the image depicting a data table;
- performing optical character recognition on the image to identify a plurality of bounding boxes and to determine one or more characters associated with each bounding box of the plurality of bounding boxes, wherein performing optical character recognition comprises:
- applying an ensemble of machine learning text detection models to identify the plurality of bounding boxes by determining coordinates for each bounding box, the ensemble of machine learning text detection models comprising:
- a first text detection model configured to detect characters;
- and
- a second text detection model configured to detect groups of characters;
- and
- applying a machine learning text recognition model to identify the one or more characters associated with each bounding box of the plurality of bounding boxes;
- labeling, using a multi-modal machine learning model trained to perform a classification task, a first bounding box of the plurality of bounding boxes as a column header and a second bounding box of the plurality of bounding boxes as including tabular data;
- and
- constructing, based at least in part on the labels of the first bounding box and the second bounding box, a digital data table corresponding to the data table.
A data extraction service comprising:
- a processor;
- and
- memory storing instructions that, when executed by the processor, causes the data extraction service to:
- receive an image depicting a data table;
- perform optical character recognition on the image to identify a plurality of bounding boxes and to determine one or more characters associated with each bounding box of the plurality of bounding boxes, wherein performing optical character recognition comprises applying a plurality of machine learning models, the plurality of machine learning models comprising a first text detection model, a second text detection model, and a text recognition model;
- label, using a multi-modal machine learning model trained to perform a classification task, a first bounding box of the plurality of bounding boxes as including a column header and a second bounding box of the plurality of bounding boxes as including tabular data;
- and
- construct, based at least in part on the labels of the first bounding box and the second bounding box, a digital data table corresponding to the data table.
A system for converting a data table into a digital data table, the system comprising:
- a device;
- a downstream entity;
- and
- a data extraction service communicatively coupled to the device and the downstream entity;
- wherein the data extraction service is configured to:
- receive an image from the device, the image depicting the data table;
- perform optical character recognition on the image to identify a plurality of bounding boxes and to determine one or more characters associated with each bounding box of the plurality of bounding boxes, wherein performing optical character recognition comprises applying a plurality of machine learning models;
- label, using a multi-modal machine learning model trained to perform a classification task, a first bounding box of the plurality of bounding boxes as a column header and a second bounding box of the plurality of bounding boxes as including tabular data;
- construct, based at least in part on the labels of the first bounding box and the second bounding box, the digital data table corresponding to the data table;
- and
- provide the digital data table to the downstream entity.
Dependencies are inferred from the claim text.