US 12,254,273 B2
Natural language processing techniques using target composite sentiment designation
Rajesh Sabapathy, Gurugram (IN); Sumeet Jain, Gurugram (IN); Saurabh Bhargava, Eden Prairie, MN (US); Sandeep Chandra Das, Bhubaneswar (IN); Gourav Awasthi, Gurugram (IN); Praveen Bansal, Hanumangarh (IN); Gaurav, New Delhi (IN); and Animesh, Patna (IN)
Assigned to Optum, Inc., Minnetonka, MN (US)
Filed by Optum, Inc., Minnetonka, MN (US)
Filed on Nov. 2, 2021, as Appl. No. 17/517,120.
Prior Publication US 2023/0137260 A1, May 4, 2023
Int. Cl. G06F 40/30 (2020.01); G06N 3/04 (2023.01)
CPC G06F 40/30 (2020.01) [G06N 3/04 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A computer-implemented method comprising:
receiving, using one or more processors, an input text sequence;
generating, using the one or more processors and responsive to inputting the input text sequence into a neural network, an intermediate emotional sentiment score object that represents the input text sequence as a plurality of intermediate emotional sentiment evaluation dimensions, each associated with a distinct emotion of a plurality of distinct emotions;
generating, using the one or more processors, based at least in part on a subset of the plurality of intermediate emotional sentiment evaluation dimensions that is associated with a target composite sentiment designation, an overall emotional sentiment score by applying respective weights to the subset of the plurality of intermediate emotional sentiment evaluation dimensions; and
until the overall emotional sentiment score satisfies one or more emotional sentiment score thresholds, iteratively modifying, using the one or more processors, the input text sequence.