US 12,093,651 B1
Machine learning techniques for natural language processing using predictive entity scoring
Nathan H. Funk, Elk River, MN (US); Eric D. Tryon, Scandia, MN (US); Amy L. Jensen, Brainerd, MN (US); Sudheer Ponnala, Chandler, AZ (US); M. P. S. Jagannadha Rao, Hyderabad (IN); Raghav Bali, Delhi (IN); Veera Raghavendra Chikka, Hyderabad (IN); Subhadip Maji, Medinipur (IN); and Anudeep Srivatsav Appe, Warangal (IN)
Assigned to Optum, Inc., Minnetonka, MN (US)
Filed by Optum, Inc., Minnetonka, MN (US)
Filed on Feb. 9, 2022, as Appl. No. 17/650,457.
Claims priority of provisional application 63/148,992, filed on Feb. 12, 2021.
Claims priority of provisional application 63/242,784, filed on Sep. 10, 2021.
Int. Cl. G06F 40/279 (2020.01); G06F 40/295 (2020.01); G06F 40/30 (2020.01); G06N 20/00 (2019.01); G06F 40/284 (2020.01)
CPC G06F 40/295 (2020.01) [G06F 40/30 (2020.01); G06N 20/00 (2019.01); G06F 40/284 (2020.01)] 20 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
identifying, by one or more processors, an entity data object associated with an input document data object using an extractive list data object associated with the entity data object and generated by:
(i) receiving, using a structured text extraction machine learning model, a structured text representation for a target document element of a training document data object,
(ii) receiving, using a structured text embedding machine learning model, an element-wise embedding based at least in part on the structured text representation for the target document element,
(iii) determining an element-wise embedding similarity measure for the element-wise embedding and a master list embedding associated with a historical extractive list data object, and
(iv) in response to determining that the element-wise embedding similarity measure satisfies an element-wise embedding similarity measure threshold, updating the historical extractive list data object with the structured text representation;
generating, by the one or more processors and using an entity scoring machine learning model, a predicted entity score for the entity data object with respect to one or more target sections of the input document data object; and
initiating, by the one or more processors, the performance of a prediction-based action based at least in part on the predicted entity score.