# US 11,907,664 B1 · Computer-based systems involving machine learning associated with generation of predictive content for data structure segments and methods of use thereof

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,907,664 B1
- Document ID: US11907664B1
- Status: Granted
- Kind: B1
- Application number: 17/245,680
- Filed: 2021-04-30
- Published: 2024-02-08
- Granted: 2024-02-20
- Earliest claimed: 2021-04-30
- Assignees: Capital One Services, LLC
- Applicants: Capital One Services, LLC
- Inventors: David Tobey; Kamalesh Jayaraman; Jagdeep Kalra
- Web page: https://patentgrep.com/patent/US11907664B1

## CPC

- G06F 40/30: Semantic analysis
- G06F 16/9535: Search customisation based on user profiles and personalisation
- G06F 40/166: Editing, e.g. inserting or deleting
- G06N 5/04: Inference or reasoning models
- G06N 20/00: Machine learning
- G06Q 10/107: Computer-aided management of electronic mailing [e-mailing]
- G06Q 10/40: Business processes related to social networking or social networking services
- G06Q 30/01: Customer relationship services
- G06Q 30/0201: Market modelling; Market analysis; Collecting market data
- G06Q 30/0271: {Personalized advertisement}
- H04L 51/046: Interoperability with other network applications or services
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M)

## Abstract

Systems and methods associated with generation and/or provision of predictive content are disclosed. One exemplary method includes receiving communications associated with a plurality of customers; determining a message type for each message of the communications; splitting first messages of the first message type into a first set of subcomponent text sections; splitting second messages of the second message type into a second set of subcomponent text sections; analyzing the first set and the second set to generate a plurality of semantic numerical scores for each respective subcomponent text section; determining at least one impactful semantic category for a target audience by selecting at least one semantic category corresponding to at least one semantic numerical score of at least one subcomponent text section of the first set or the second set that is equal to or higher than a first pre-determined threshold value; generating personalized textual content targeting the audience.

## Claims

### Claim 1

A computer-implemented method comprising:
training, by at least one processor, a natural language generation (NLG) machine learning model on customer data derived from a plurality of customers;
receiving, by the at least one processor, communications associated with the plurality of customers, the communications comprised a plurality of customer messages;
determining, by the at least one processor, a message type, from among a plurality of message types, for each message of the communications,
wherein the plurality of message types comprises a first message type and a second message type;
wherein the communications comprise the first messages of the first message type and second messages of the second message type
splitting, by the at least one processor, the first messages of the first message type into a first set of subcomponent text sections;
splitting, by the at least one processor, the second messages of the second message type into a second set of subcomponent text sections;
predicting, based on the plurality of customer messages in combination with conversion event data pertaining to the plurality of customer messages, a set of predicted customer preferences for the plurality of customers;
analyzing, by the at least one processor, via the NLG machine learning model, the first set and the second set in combination with the set of predicted customer preferences to generate a plurality of semantic numerical scores comprising a predicted effectiveness for each respective subcomponent text section, wherein each respective semantic numerical score is based on an evaluation of each respective subcomponent text section in a respective semantic category of a plurality of semantic categories and the predicted effectiveness comprises predicted conversion rates for messages that follow the set of predicted customer preferences,
wherein the plurality of the semantic categories comprises at least three semantic categories are selected from a sentiment category, an emotion category, a perceived message type category, a semantic relatedness category, a feeling category, a tone category, a perception category, a micro structure category, and an emotional intelligence category;
determining, by the at least one processor, at least one impactful semantic category for a target audience by selecting at least one semantic category corresponding to at least one semantic numerical score of at least one subcomponent text section of the first set or the second set that is equal to or higher than a first pre-determined threshold value;
generating, by the at least one processor, via the NLG machine learning model, personalized textual content targeting the audience based on at least one unit of text having a corresponding semantic numerical score in the at least one impactful semantic category that is equal to or higher than a second pre-determined threshold value; and
generating, by the at least one processor, via the NLG machine learning model, at least one personalized communication for transmission to the audience from the personalized textual content; and
updating, by the at least one processor, the NLG machine learning model based on identified interactions between the audience and the personalized communication.

### Claim 2 (depends on 1)

The method of claim 1 wherein the first message type comprises emails and the second message type comprises SMS messages, push messages, and web banners.

### Claim 3 (depends on 1)

The method of claim 1 wherein the first set of subcomponent text sections comprises 3 or more parts, including 3 or more of a subject line, a preheader, a banner image, an introductory section, and a call to action.

### Claim 4 (depends on 1)

The method of claim 1 wherein the first message type comprises email messages, and wherein the first set of subcomponent text sections comprises 3 or more parts, including 3 or more of a subject line, a preheader, a banner image, an introductory section, and a call to action.

### Claim 5 (depends on 1)

The method of claim 1 wherein the second set of subcomponent text sections comprises 2 or more parts, including 2 or more of an introductory section, a body section, a value proposition, and an end section.

### Claim 6 (depends on 1)

The method of claim 1 wherein the second message type comprises 2 or more of SMS messages, push messages and/or web banners, and wherein the second set of subcomponent text sections comprises 2 or more parts, including 2 or more of an introductory section, a body section, a value proposition, an end section, and/or entire message.

### Claim 7 (depends on 1)

The method of claim 1 wherein the sentiment category is comprised of 2 or more subcategories selected from positive, neutral, and negative.

### Claim 8 (depends on 1)

The method of claim 1 wherein the emotion category is comprised of 2 or more, subcategories selected from a group composed of happiness, concern, excitement, sadness, candor and boredom.

### Claim 9 (depends on 1)

The method of claim 1 wherein the perceived message type category is comprised of 2 or more subcategories selected from a group composed of news, feedback, query, marketing, and spam.

### Claim 10 (depends on 1)

The method of claim 1 wherein the at least one personalized communication comprises a first portion of the personalized textual content that corresponds to the sentiment category determined to be impactful to the audience, a second portion of the personalized textual content that corresponds to the emotion category determined to be impactful to the audience, and a third portion of the personalized textual content that corresponds to the perceived message type category determined to be impactful to the audience.

### Claim 11 (depends on 1)

The method of claim 1 wherein the semantic numerical score in the semantic relatedness category is set based on a percentage of the text that is determined to be semantically similar to a benchmark communication.

### Claim 12 (depends on 1)

The method of claim 1 further comprising:
building a database of information regarding personalized messages that are impactful to each individual customer, wherein the database is built based on at least one prior personalized message that elicited a response from the individual customer.

### Claim 13 (depends on 1)

The method of claim 1 further comprising:
analyzing different textual components or text within at least one region of the at least one personalized communication to determine how the different textual component or text affect the semantic numerical scores; and/or
changing the different textual portions or text within the at least one region of the at least one personalized communication, prior to sending to the audience, to generate a personalized communication that is determined to be potentially impactful to the audience via an increase in the semantic numerical score.

### Claim 14

A computer-implemented method comprising:
training, by at least one processor, a natural language generation (NLG) machine learning model on customer data derived from a plurality of customers;
receiving, by the at least one processor, communications associated with the plurality of customers, the communications comprised of customer messages;
categorizing the customer messages that are emails as a first message type, and categorizing the customer messages that are SMS messages, push messages, and web banners as a second message type;
splitting, by the at least one processor, first messages of the first message type into a first set of subcomponent text sections;
splitting, by the at least one processor, second messages of the second message type into a second set of subcomponent text sections;
predicting, based on the plurality of customer messages in combination with conversion event data pertaining to the plurality of customer messages, a set of predicted customer preferences for the plurality of customers;
analyzing, by the at least one processor, via the NLG machine learning model, the first set and the second set in combination with the set of predicted customer preferences to generate a plurality of semantic numerical scores comprising a predicted effectiveness for each respective subcomponent text section, wherein each respective semantic numerical score is based on an evaluation of each respective subcomponent text section in a respective semantic category of a plurality of semantic categories and the predicted effectiveness comprises predicted conversion rates for messages that follow the set of predicted customer preferences,
wherein the plurality of the semantic categories comprises at least three semantic categories are selected from a sentiment category, an emotion category, a perceived message type category, a semantic relatedness category, a feeling category, a tone category, a perception category, a micro structure category, and an emotional intelligence category;
determining, by the at least one processor, at least one impactful semantic category for a target audience by selecting at least one semantic category corresponding to at least one semantic numerical score of at least one subcomponent text section of the first set or the second set that is equal to or higher than a first pre-determined threshold value;
generating, by the at least one processor, via the NLG machine learning model, personalized textual content targeting the audience based on at least one unit of text having a corresponding semantic numerical score in the at least one impactful semantic category that is equal to or higher than a second pre-determined threshold value;
generating, by the at least one processor, via the NLG machine learning model, at least one personalized communication for transmission to the audience from the personalized textual content; and
updating, by the at least one processor, the NLG machine learning model based on identified interactions between the audience and the personalized communication.

### Claim 15 (depends on 14)

The method of claim 14 wherein the first set of subcomponent text sections comprises 3 or more parts, including 3 or more of a subject line, a preheader, a banner image, an introductory section, and a call to action.

### Claim 16 (depends on 14)

The method of claim 14 wherein the first message type comprises email messages, and wherein the first set of subcomponent text sections comprises 3 or more parts, including 3 or more of a subject line, a preheader, a banner image, an introductory section, and a call to action.

### Claim 17 (depends on 14)

The method of claim 14 wherein the second set of subcomponent text sections comprises 2 or more parts, including 2 or more of an introductory section, a body section, a value proposition, and an end section.

### Claim 18 (depends on 14)

The method of claim 14 wherein the second message type comprises 2 or more of SMS messages, push messages and/or web banners, and wherein the second set of subcomponent text sections comprises 2 or more parts, including 2 or more of an introductory section, a body section, a value proposition, an end section, and/or entire message.

### Claim 19 (depends on 14)

The method of claim 14 wherein the sentiment category is comprised of 2 or more subcategories selected from positive, neutral, and negative.

### Claim 20 (depends on 14)

The method of claim 14 wherein the emotion category is comprised of 3 or more subcategories selected from a group composed of happiness, concern, excitement, sadness, candor and boredom.