Using machine learning for individual classification

Patent metadata

Field	Value
Publication number	US-12430513-B2
Application number	US-202217722566-A
Country	US
Kind code	B2
Filing date	Apr 18, 2022
Priority date	Apr 18, 2022
Publication date	Sep 30, 2025
Grant date	Sep 30, 2025

Abstract

A method comprises analyzing a plurality of natural language inputs associated with at least one user, and determining a plurality of contexts for the plurality of natural language inputs based, at least in part, on the analysis. In the method, a plurality of relationships linked to the at least one user are identified based, at least in part, on the analysis, and the at least one user is classified in one or more categories based, at least in part, on the plurality of contexts and the plurality of relationships. At least one of the analyzing, determining, identifying and classifying is performed using one or more machine learning models.

First claim

What is claimed is:

1. A method, comprising:
capturing, using one or more virtual agents, a plurality of communications associated with at least one user;
generating a plurality of natural language inputs based on the plurality of communications associated with the at least one user using one or more machine learning algorithms;
analyzing the plurality of natural language inputs associated with the at least one user using the one or more machine learning algorithms, the analyzing comprising:
determining a plurality of contexts for the plurality of natural language inputs based on one or more natural language processing algorithms of the one or more machine learning algorithms;
identifying a plurality of relationships linked to the at least one user based on one or more other users associated with the plurality of communications, and one or more domains in which the plurality of communications occurred;

generating a corpus associated with the at least one user comprising intent corpus data and sentiment corpus data based on the determined plurality of contexts, words and phrases associated with the plurality of natural language inputs, and relationship corpus data based on the identified plurality of relationships;
training a first machine learning model using the of the intent corpus data of the generated corpus associated with the at least one user, the input corpus data comprising intents corresponding to the words and phrases associated with the plurality of natural language inputs based on the plurality of communications associated with the at least one user;
training a second machine learning model using the of the sentiment corpus data of the generated corpus associated with the at least one user, the sentiment corpus data comprising sentiments corresponding to the words and phrases associated with the plurality of natural language inputs based on the plurality of communications associated with the at least one user;
training a third machine learning model using the of the relationship corpus data of the generated corpus associated with the at least one user, the relationship corpus data comprising interactions between the at least one user and the one or more other users associated with the plurality of communications, one or more actions taken by the at least one user and the one or more other users in the interactions, and domains of the one or more domains in which the interactions and actions occurred, based on the identified plurality of relationships linked to the at least one user;
predicting, with the trained first machine learning model, a plurality of intents for a plurality of incoming communications associated with the at least one user;
predicting, with the trained second machine learning model, a plurality of sentiments for the plurality of incoming communications associated with the at least one user and a priority for the plurality of incoming communications associated with the at least one user based on the plurality of sentiments;
generating, with the trained third machine learning model, one or more relationship graphs associated with the identified plurality of relationships linked to the at least one user in the plurality of incoming communications;
saving the plurality of intents, plurality of sentiments and the one or more relationship graphs to an individual profile associated with the at least one user;
training a fourth machine learning model using the plurality of intents, the plurality of sentiments, and the one or more relationship graphs of the individual profile associated with the at least one user;
classifying the at least one user, using the trained fourth machine learning model, in one or more categories based, at least in part, on the plurality of intents, the plurality of sentiments, and the one or more relationship graphs of the individual profile associated with the at least one user;
generating one or more recommendations for the at least one user based on the classification of the at least one user into the one or more categories, using the fourth machine learning model; and
iteratively updating the individual profile associated with the at least one user by:
inputting one or more additional communications associated with the at least one user into the first, second and third machine learning models to generate an updated plurality of intents, an updated plurality of sentiments, and an updated one or more relationship graphs;
updating the individual profile associated with the at least one user with the updated plurality of intents, the updated plurality of sentiments, and the updated one or more relationship graphs;
inputting the updated plurality of intents, the updated plurality of sentiments, and the updated one or more relationship graphs of the updated individual profile associated with the at least one user into the fourth machine learning model to adjust the fourth machine learning model;
re-classifying the at least one user into the one or more categories with the adjusted fourth machine learning model based, at least in part, on the updated plurality of sentiments, and the updated one or more relationship graphs of the updated individual profile associated with the at least one user; and
generating one or more additional recommendations for the at least one user based on the classification of the at least one user into the one or more categories;

wherein the steps of the method are executed by a processing device operatively coupled to a memory.

2. The method of claim 1 , wherein the first machine learning model comprises a bi-directional recurrent neural network.

3. The method of claim 1 , wherein the second machine learning model comprises a bi-directional recurrent neural network.

4. The method of claim 1 , wherein the one or more relationship graphs include the plurality of relationships modeled by a plurality of nodes, wherein the plurality of relationships between the plurality of nodes comprise edges of the one or more relationship graphs.

5. The method of claim 4 , wherein the plurality of nodes comprise at least one of one or more persons, the one or more domains, one or more sub-domains, one or more functions, one or more utilities and one or more activities.

6. The method of claim 4 , wherein the relationships between the plurality of nodes comprise interactions between respective pairs of the plurality of nodes.

7. The method of claim 4 , wherein the one or more relationship graphs are in one of a resource description framework (RDF) format and a labeled property graph (LPG) format.

8. The method of claim 4 , further comprising extracting a plurality of features from the plurality of intents, the plurality of sentiments and from the one or more relationship graphs, wherein classifying the at least one user into the one or more categories comprises inputting the plurality of features to a neural network of the fourth machine learning model which predicts the one or more categories for the at least one user.

9. The method of claim 8 , further comprising training the neural network of the fourth machine learning model with context data and relationship data.

10. The method of claim 8 , wherein the neural network of the fourth machine learning model comprises at least two hidden layers utilizing a rectified linear unit activation function.

11. The method of claim 8 , wherein the neural network of the fourth machine learning model comprises a plurality of nodes connected with each other, wherein respective ones of connections comprise a weight factor and respective ones of the plurality of nodes comprise a bias factor.

12. The method of claim 1 , wherein analyzing the plurality of natural languag

Assignees

• Dell Products Lp

Inventors

• Kumar Dhilip S
• Mohanty Bijan Kumar
• Dinh Hung
• Sahoo Sujit Kumar

Classifications

• G10L15/1822

  Parsing for meaning understanding · CPC title

• G10L25/30

  using neural networks · CPC title

• G06F40/35

  Discourse or dialogue representation · CPC title

• G10L15/1807

  using prosody or stress · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title