US 12,321,703 B2
Generative model for aspect term extraction
Amir Pouran Ben Veyseh, Eugene, OR (US); and Franck Dernoncourt, San Jose, CA (US)
Assigned to ADOBE INC., San Jose, CA (US)
Filed by ADOBE INC., San Jose, CA (US)
Filed on Feb. 7, 2022, as Appl. No. 17/650,177.
Prior Publication US 2023/0252237 A1, Aug. 10, 2023
Int. Cl. G06F 40/289 (2020.01); G06F 40/35 (2020.01); G06N 3/08 (2023.01)
CPC G06F 40/289 (2020.01) [G06F 40/35 (2020.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for natural language processing, comprising:
receiving an input phrase including an aspect term;
generating, using a language generator model, a complement phrase including the aspect term based on the input phrase, wherein the complement phrase includes different words than the input phrase that represent complementary information from the input phrase based on a regularization loss that minimizes a similarity between the input phrase and the complement phrase based on a dot product between representation vectors;
combining the input phrase and the complement phrase into a word embedding model to obtain an augmented representation of the input phrase by encoding the input phrase and the complement phrase to obtain an input representation and a complement representation, respectively, in a vector embedding space and combining the input representation and the complement representation to obtain the augmented representation; and
generating sentiment information corresponding to the aspect term based on the augmented representation.