US 11,657,104 B2
Scalable ground truth disambiguation
Faheem Altaf, Pflugerville, TX (US); Lisa Seacat Deluca, Baltimore, MD (US); and Raghuram Srinivas, McKinney, TX (US)
Filed by International Business Machines Corporation, Armonk, NY (US)
Filed on Oct. 21, 2019, as Appl. No. 16/659,165.
Application 16/659,165 is a continuation of application No. 15/490,081, filed on Apr. 18, 2017, granted, now 10,572,826.
Prior Publication US 2020/0050969 A1, Feb. 13, 2020
This patent is subject to a terminal disclaimer.
Int. Cl. G06F 16/953 (2019.01); G06N 20/00 (2019.01); G06F 40/20 (2020.01); G06N 5/025 (2023.01)
CPC G06F 16/953 (2019.01) [G06F 40/20 (2020.01); G06N 5/025 (2013.01); G06N 20/00 (2019.01)] 20 Claims
OG exemplary drawing
1. A computer implemented method comprising:
obtaining an utterance input from a user agent;
collecting context data of the utterance input from the user agent, wherein the context data describes circumstances of the utterance input;
generating a context tag based on the context data, wherein the context tag corresponds to the utterance input;
selecting a ground truth, wherein the selecting includes using the utterance input and the context tag, wherein the ground truth includes an utterance and an intent, wherein the utterance of the ground truth is semantically identical to the utterance input, and wherein the intent of the ground truth is semantically consistent with the context tag; and
updating the ground truth, wherein the updating includes attaching the context tag, wherein the updating the ground truth includes updating a first ground truth so that the first ground truth includes the context tag, and training a machine learning process using first training data, wherein the first training data used to train the machine learning process includes the first ground truth tagged with the context tag and having a first utterance and a first intent.