US 12,072,919 B2
Sentence level dialogue summaries using unsupervised machine learning for keyword selection and scoring
Nikita Alekseyevich Lukyanenko, San Jose, CA (US); and Alexander Shvid, Las Vegas, NV (US)
Assigned to PAYPAL, INC., San Jose, CA (US)
Filed by PAYPAL, INC., San Jose, CA (US)
Filed on Aug. 31, 2021, as Appl. No. 17/463,303.
Prior Publication US 2023/0063713 A1, Mar. 2, 2023
Int. Cl. G06F 16/30 (2019.01); G06F 16/34 (2019.01); H04L 51/02 (2022.01); H04L 51/046 (2022.01)
CPC G06F 16/345 (2019.01) [H04L 51/02 (2013.01); H04L 51/046 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising:
a non-transitory memory; and
one or more hardware processors coupled to the non-transitory memory and configured to read instructions from the non-transitory memory to cause the system to perform operations comprising:
receiving a dialogue between a user and a live chat service of a service provider associated with the system, wherein the dialogue comprises one or more asynchronous chat sessions between the user and the live chat service, and wherein the one or more asynchronous chat sessions comprise a plurality of sentences;
processing the dialogue using a machine learning (ML) model pipeline comprising a plurality of ML models, wherein the processing comprises:
performing a keyword level selection for the dialogue using the plurality of sentences and a first one of the plurality of ML models a machine learning (ML) model, and
determining a plurality of keywords in the plurality of sentences and a corresponding ranking of each of the plurality of keywords using a second one of the plurality of ML models and based on the keyword level selection;
calculating, by the ML model pipeline, scores for the plurality of sentences based on a number of appearances of the plurality of keywords in the plurality of sentences and the corresponding ranking for each of the plurality of keywords;
computing modifications to the scores based on a particular agent and feedback by the particular agent from sentence summarizations of previous dialogues; and
outputting a subset of the plurality of sentences based on the scores and the modifications.