# US 2019/0377785 A1 · Automatically generating digital enterprise content variants

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

- Number: US 2019/0377785 A1
- Document ID: US20190377785A1
- Status: Published application
- Kind: A1
- Application number: 16/005,217
- Filed: 2018-06-11
- Published: 2019-12-12
- Earliest claimed: 2018-06-11
- Assignees: Adobe Inc.
- Applicants: Adobe Inc.
- Inventors: Anandhavelu N; Padmanabhan Anandan; Niyati Chhaya; Cedric Huesler; Balaji Vasan Srinivasan; Atanu R Sinha
- Web page: https://patentgrep.com/patent/US20190377785A1

## CPC

- G06F 40/131: Fragmentation of text files, e.g. creating reusable text-blocks; Linking to fragments, e.g. using XInclude; Namespaces
- G06F 3/0482: Interaction with lists of selectable items, e.g. menus
- G06F 40/166: Editing, e.g. inserting or deleting
- G06F 40/20: Natural language analysis (semantic analysis of natural language G06F40/30)
- G06F 40/253: Grammatical analysis; Style critique
- G06N 20/00: Machine learning
- G06F 40/56: Natural language generation
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

This disclosure relates to methods, non-transitory computer readable media, and systems that, based on a sparse textual segment, can use machine learning models to generate document variants that are both conforming to digital content guidelines and uniquely tailored for distribution to client devices of specific audiences via specific delivery channels. To create such variants, in some embodiments, the methods, non-transitory computer readable media, and systems generate suggested modifications to a draft document that correspond to features of content-guideline-conforming documents. Additionally, or alternatively, in certain implementations, the disclosed methods, non-transitory computer readable media, and systems generate suggested modifications to a draft document that correspond to features of audience-channel-specific documents.

## Claims

### Claim 1

In a digital medium environment for generating and implementing digital content campaigns across computer networks, a computer-implemented method for machine generating digital-enterprise-content variants from textual segments, comprising:
receiving, from a publisher device, user input indicating a textual segment;
composing a draft document from digital-content-campaign documents identified from a query based on the textual segment;
performing a step for proposing suggested modifications to the draft document based on content-guideline-conforming documents and audience-channel-specific documents;
providing, to the publisher device, selectable options corresponding to the suggested modifications to the draft document; and
revising the draft document to incorporate one or more modifications of the suggested modifications based on a user selection of one or more selectable options from the selectable options.

### Claim 2 (depends on 1)

The method of claim 1, wherein the suggested modifications to the draft document comprise a modification to the draft document that reduces a difference in an attribute dimension score for the draft document and a representative attribute dimension score for the content-guideline-conforming documents.

### Claim 3 (depends on 2)

The method of claim 2, wherein providing the selectable options corresponding to the suggested modifications to the draft document comprises:
providing, for display on the publisher device, a user interface comprising the one or more selectable options and an attribute-dimension-score element comprising both the attribute dimension score for the draft document and the representative attribute dimension score for the content-guideline-conforming documents.

### Claim 4 (depends on 1)

The method of claim 1, wherein the audience-channel-specific documents reflect documents corresponding to a selected audience or reflect documents corresponding to a selected delivery channel.

### Claim 5 (depends on 1)

The method of claim 1, wherein the suggested modifications to the draft document comprise a modification that reduces a difference between a communication tone score for the draft document and a representative communication tone score for the audience-channel-specific documents.

### Claim 6 (depends on 1)

The method of claim 1, wherein composing the draft document from the digital-content-campaign documents identified from the query comprises:
extracting key words from the textual segment based on a term frequency of words from the textual segment and an inverse document frequency of the words in the digital-content-campaign documents;
querying the digital-content-campaign documents for the key words to identify content-campaign-document fragments; and
composing the draft document by mapping words from the content-campaign-document fragments to a graph and generating candidate sentences for the draft document based on weighted word nodes within the graph that reflect the query.

### Claim 7

A non-transitory computer readable storage medium comprising instructions that, when executed by at least one processor, cause a system to:
in response to identifying user input from a publisher device indicating a textual segment, compose a draft document from digital-content-campaign documents identified from a query based on the textual segment;
generate a first suggested modification to the draft document, for display on the publisher device, by comparing the draft document with content-guideline-conforming documents;
in response to identifying user input from the publisher device indicating at least one of an audience or a delivery channel, generate a second suggested modification to the draft document, for display on the publisher device, by comparing the draft document with audience-channel-specific documents; and
based on a user selection of a modification from at least one of the first suggested modification or the second suggested modification, modify the draft document to incorporate the modification from the first suggested modification and the second suggested modification.

### Claim 8 (depends on 7)

The non-transitory computer readable storage medium of claim 7, further comprising instructions that, when executed by the at least one processor, cause the system to generate the first suggested modification to the draft document by:
utilizing a machine learning model to determine an attribute dimension score for the draft document based on textual features of the draft document;
utilizing the machine learning model to determine an attribute-dimension-score distribution for the content-guideline-conforming documents based on textual features of the content-guideline-conforming documents; and
comparing the attribute dimension score for the draft document and the attribute-dimension-score distribution for the content-guideline-conforming documents.

### Claim 9 (depends on 8)

The non-transitory computer readable storage medium of claim 8, further comprising instructions that, when executed by the at least one processor, cause the system to generate the first suggested modification to the draft document by determining that the first suggested modification reduces a difference between the attribute dimension score for the draft document and the attribute-dimension-score distribution for the content-guideline-conforming documents.

### Claim 10 (depends on 8)

The non-transitory computer readable storage medium of claim 8, further comprising instructions that, when executed by the at least one processor, cause the system to generate the first suggested modification to the draft document by providing, for display on the publisher device, a user interface comprising the first suggested modification and an attribute-dimension-score element comprising both the attribute dimension score for the draft document and the attribute-dimension-score distribution for the content-guideline-conforming documents.

### Claim 11 (depends on 7)

The non-transitory computer readable storage medium of claim 7, further comprising instructions that, when executed by the at least one processor, cause the system to generate the second suggested modification to the draft document for distribution to the audience by:
identifying the audience-channel-specific documents by collecting digital documents distributed to client devices corresponding to the audience; and
utilizing a machine learning model to determine a communication-tone-score distribution for the audience-channel-specific documents.

### Claim 12 (depends on 11)

The non-transitory computer readable storage medium of claim 11, further comprising instructions that, when executed by the at least one processor, cause the system to generate the second suggested modification to the draft document for distribution to the audience by:
utilizing a machine learning model to determine a communication tone score for the draft document; and
comparing the communication tone score for the draft document with the communication-tone-score distribution for the audience-channel-specific documents.

### Claim 13 (depends on 7)

The non-transitory computer readable storage medium of claim 7, further comprising instructions that, when executed by the at least one processor, cause the system to generate the second suggested modification to the draft document for distribution via the delivery channel by:
identifying the audience-channel-specific documents by collecting digital documents distributed to client devices via the delivery channel; and
comparing a communication tone score for the draft document with a communication-tone-score distribution for the audience-channel-specific documents.

### Claim 14 (depends on 13)

The non-transitory computer readable storage medium of claim 13, further comprising instructions that, when executed by the at least one processor, cause the system to determine communication tone scores indicating a degree to which the audience-channel-specific documents exhibit a tone dimension based on communication tone features of the audience-channel-specific documents, wherein:
the tone dimension comprises at least one of a formal tone, a frustrated tone, or a polite tone; and
the communication tone features of the draft document comprise at least one of psycholinguistic features, lexical features, syntactic features, derived features, or formality features of the draft document.

### Claim 15

A system for machine generating digital-enterprise-content variants from textual segments comprising:
at least one processor;
at least one non-transitory computer memory comprising content-guideline-conforming documents, audience-channel-specific documents, and instructions that, when executed by at least one processor, cause the system to:
generate, for display on a publisher device, a first suggested modification to a draft document by:
utilizing a machine learning model to determine an attribute dimension score for the draft document based on textual features of the draft document and an attribute-dimension-score distribution for the content-guideline-conforming documents based on textual features of the content-guideline-conforming documents; and
generating the first suggested modification based on a comparison of the attribute dimension score and the attribute-dimension-score distribution; and
generate, for display on the publisher device, a second suggested modification to the draft document by comparing a communication tone score for the draft document with communication tone scores for the audience-channel-specific documents; and
based on a user selection of a modification from at least one of the first suggested modification or the second suggested modification, modify the draft document to incorporate the modification from the first suggested modification and the second suggested modification.

### Claim 16 (depends on 15)

The system of claim 15, wherein the textual features of the draft document comprise one or more of affect-based features, linguistic features, positive-negative-term features, term features, or term frequency features of the draft document.

### Claim 17 (depends on 15)

The system of claim 15, further comprising instructions that, when executed by the at least one processor, cause the system to:
modify the draft document to incorporate the first suggested modification based on an indication from the publisher device of a first selection by a user of the first suggested modification;
generate the second suggested modification to the draft document based on the first selection of the first suggested modification; and
modify the draft document to incorporate the second suggested modification based on an indication from the publisher device of a second selection by the user of the second suggested modification.

### Claim 18 (depends on 15)

The system of claim 15, further comprising instructions that, when executed by the at least one processor, cause the system to utilize the machine learning model to determine the attribute dimension score and the attribute-dimension-score distribution by:
utilizing the machine learning model to determine attribute dimension scores indicating a degree to which the content-guideline-conforming documents exhibit an attribute dimension based on the textual features of the content-guideline-conforming documents;
determining the attribute-dimension-score distribution for the content-guideline-conforming documents based on the attribute dimension scores; and
utilizing the machine learning model to determine the attribute dimension score indicating a degree to which the draft document exhibits the attribute dimension based on textual features of the draft document.

### Claim 19 (depends on 18)

The system of claim 18, wherein the attribute dimension comprises a competence dimension, an excitement dimension, a ruggedness dimension, a sincerity dimension, or a sophistication dimension.

### Claim 20 (depends on 15)

The system of claim 15, further comprising instructions that, when executed by the at least one processor, cause the system to generate the first suggested modification by:
determining that the attribute-dimension-score distribution does not include the attribute dimension score for the draft document; and
suggesting a modification to a word or phrase within the draft document based on determining that the attribute-dimension-score distribution does not include the attribute dimension score for the draft document.
