US 12,229,208 B2
Responsive category prediction for user queries
Ali Ahmadvand, Atlanta, GA (US); Surya Kallumadi, Atlanta, GA (US); and Faizan Javed, Atlanta, GA (US)
Assigned to Home Depot Product Authority, LLC, Atlanta, GA (US)
Filed by Home Depot Product Authority, LLC, Atlanta, GA (US)
Filed on Sep. 28, 2021, as Appl. No. 17/487,732.
Claims priority of provisional application 63/085,518, filed on Sep. 30, 2020.
Prior Publication US 2022/0100806 A1, Mar. 31, 2022
Int. Cl. G06F 16/954 (2019.01); G06F 18/214 (2023.01); G06F 40/205 (2020.01); G06N 3/04 (2023.01)
CPC G06F 16/954 (2019.01) [G06F 18/2148 (2023.01); G06F 40/205 (2020.01); G06N 3/04 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A computer-implemented method for determining a category responsive to a user query, the method comprising:
receiving, by a computing device, a training data set comprising a plurality of data pairs, each data pair comprising: (i) a query; and (ii) an associated one or more categories that are responsive to the query, wherein the one or more categories in the training data set defines a plurality of categories;
training, by the computing device, a machine learning algorithm, according to the training data set, to create a trained model, wherein training the machine learning algorithm comprises:
separating, by the computing device, each query into a respective one or more words that comprise the query;
calculating, by the computing device, respective embeddings corresponding to vectors for each of the one or more words to create a word embeddings set;
creating, by the computing device, a first co-occurrence data structure defining co-occurrence of respective word representations of the queries with the plurality of categories based on the word embeddings set;
creating, by the computing device, a second co-occurrence data structure defining co-occurrence of respective categories in respective data pairs based on the word embeddings set;
inputting, by the computing device, the first co-occurrence data structure to a self-attention mechanism, wherein the self-attention mechanism outputs respective attention vectors representative of each respective word representations contribution to its association with each category; and
applying, by the computing device, the respective attention vectors indicative of a relative correlation between each category and each word representation as a weight set to the word embeddings set; and
deploying the trained model to return one or more categories in response to a new query input.