Automated indexing and extraction of information in digital records
AssigneeVelocityEHS Holdings, Inc.
InventorsJulia Penfield, Aatish Suman, Veeru Talreja, Misbah Zahid Khan
Life of this patent
Estimating the term…
Abstract
Systems and methods for automated indexing and extraction of information in digital documents are disclosed. A method may comprise identifying a page containing targeted information; inputting an image of the page into a visual machine learning network (visual ML), wherein the visual ML is trained to recognize text associated with the targeted information in an image; identifying by the visual ML, a section of the image that contains the targeted information; inputting the digital document, and coordinates of the section into an extraction module; and extracting the targeted information by the extraction module from the section.
Claims20
A computer implemented method to automatically index targeted information in digital records, the method comprising:
- inputting an image of a page of a digital record into a visual machine learning network (a visual ML), wherein the visual ML is trained to recognize text associated with targeted information in the image;
- identifying by the visual ML, a section of the image that contains the targeted information;
- inputting a page number of the page, the digital record, and coordinates of the section into an extraction module;
- and
- extracting the targeted information by the extraction module from the section.
A system comprising:
- at least one processor;
- and
- at least one non-transitory, computer-readable memory storing instructions that, when executed by the at least one processor, are effective to:
- input an image of a page of a digital record into a visual machine learning network (a visual ML), wherein the visual ML is trained to recognize text associated with targeted information in the image;
- identify by the visual ML, sections of the image that contain the targeted information;
- input a page number of the page, the digital record, and coordinates of the sections into an extraction module;
- extract the targeted information by the extraction module from the sections;
- input the extracted targeted information into a natural language processing ML network (an NLP);
- and
- identify at least one data item, by the NLP, based on a structure of the extracted targeted information.
A method to train machine learning networks to autonomously identify targeted information, the method comprising:
- parsing a custom dataset containing texts of digital records to produce a string of characters;
- identifying portions that contain information in the string of characters;
- splitting the string of characters into tokens;
- training a natural language processing ML network (an NLP), the training comprising:
- inputting the tokens into the NLP;
- and
- outputting, by the NLP, identifications comprising a first word of a chemical ingredient name, a subsequent word of a chemical ingredient name, or a word not belonging to any chemical ingredient name.
Dependencies are inferred from the claim text.