# US 11,308,278 B2 · Predicting style breaches within textual content

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

- Number: US 11,308,278 B2
- Document ID: US11308278B2
- Status: Granted
- Kind: B2
- Application number: 16/842,280
- Filed: 2020-04-07
- Published: 2020-08-06
- Granted: 2022-04-19
- Earliest claimed: 2017-11-14
- Assignees: Adobe Inc.
- Applicants: Adobe Inc.
- Inventors: Pranav Ravindra Maneriker; Anandhavelu Natarajan; Vivek Gupta; Basava Raj K
- Web page: https://patentgrep.com/patent/US11308278B2

## CPC

- G06F 40/253: Grammatical analysis; Style critique
- G06N 3/08: Learning methods
- G06F 40/169: Annotation, e.g. comment data or footnotes
- G06F 40/284: Lexical analysis, e.g. tokenisation or collocates
- G06F 40/289: Phrasal analysis, e.g. finite state techniques or chunking
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/09: Supervised learning
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 20/00: Machine learning
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Embodiments of the present invention provide systems, methods, and computer storage media directed to facilitating predicting style breaches within content. In one embodiment, target content for which style breach prediction is desired is obtained. Style features associated with the target content are identified. Such style features and a style breach prediction model are used to predict a style breach within the target content, the style breach indicating a change of style used within the target content (e.g., a single document).

## Claims

### Claim 1

A computerized method to facilitate style breach prediction, the method comprising:
identifying a set of style features associated with a particular content segment within a target content for which style breach detection is desired;
based on the set of style features for the particular content segment, obtaining a binary value associated with the particular content segment to determine whether a style breach between the particular content segment and a previous content segments exists; and
predicting a style breach indicating a change in writing style of text between the particular content segment and the previous content segment based on a determination that a binary value associated with the particular content segment is different from a binary value associated with the previous content segment.

### Claim 2 (depends on 1)

The method of claim 1, wherein the style features comprise lexical features.

### Claim 3 (depends on 1)

The method of claim 1, wherein the style features comprise syntactic features.

### Claim 4 (depends on 2)

The method of claim 2, wherein the lexical features comprise one or more of an average word length, a sentence length, a word length frequency, a Flesh-Kincaid readability score, a frequency of words not in the English dictionary, an Honore's Index value, a Hapax legomena value, a Hapax dislegomena value, a Yule's Index value, and a token ratio.

### Claim 5 (depends on 3)

The method of claim 3, wherein the syntactic features comprise one or more of a count of a function word, a punctuation count, an upper case count, and a part-of-speech n-gram.

### Claim 6 (depends on 1)

The method of claim 1, further comprising generating a style vector for the particular content segment based on the set of style features associated with the particular content segment.

### Claim 7 (depends on 6)

The method of claim 6, further comprising reducing dimensionality of the style features using a Principal Component analysis.

### Claim 8 (depends on 1)

The method of claim 1, wherein a style breach prediction model comprises a long short-term memory model.

### Claim 9 (depends on 1)

The method of claim 1, further comprising providing a representation of the style breach within the target content.

### Claim 10 (depends on 1)

The method of claim 1, further comprising:
determining a modification suggestion to modify the target content for style consistency; and
providing the modification suggestion.

### Claim 11

One or more non-transient computer storage media storing computer-useable instructions that, when used by one or more computing devices, cause the one or more computing devices to perform operations comprising:
identifying style features associated with a content segment of a plurality of content segments of training content;
for the content segment, generating a vector output based on a style breach annotation and the style features associated with the content segment, wherein the vector output comprises a binary value that, when different from a binary value associated with a previous content segment, indicates a style breach; and
training a style breach prediction model based on the style features and the vector outputs generated for the content segment, the style breach prediction model configured to predict changes in text style within content.

### Claim 12 (depends on 11)

The one or more non-transient computer storage media of claim 11, wherein the training content comprises multi-authored content.

### Claim 13 (depends on 11)

The one or more non-transient computer storage media of claim 11, further comprising generating a plurality of style vectors for a plurality of training content segments using the style features.

### Claim 14 (depends on 13)

The one or more non-transient computer storage media of claim 13, further comprising using a plurality of vectors and a plurality of style breach annotations to generate output values associated with each of the plurality of vectors, the output values indicating style breaches.

### Claim 15 (depends on 14)

The one or more non-transient computer storage media of claim 14, wherein the plurality of vectors and corresponding output values are used as input to train the style breach prediction model.

### Claim 16 (depends on 11)

The one or more non-transient computer storage media of claim 11, wherein the style breach prediction model comprises a long short-term memory model.

### Claim 17

A computer system comprising:
one or more processors; and
one or more non-transitory computer-readable storage media, coupled with the one or more processors, having instructions stored thereon, which, when executed by the one or more processors, cause the computing system to provide:
a style breach prediction means configured to predict a style breach between a particular content segment and a previous content segment based on a determination that a binary value associated with the particular content segment is different from a binary value associated with the previous content segment.

### Claim 18 (depends on 17)

The system of claim 17, further comprising a model generating means configured to generate a style breach prediction model using style features identified within training content and corresponding style breach annotations, wherein the model generating means is further configured to identify style features associated with the training content.

### Claim 19 (depends on 18)

The system of claim 18, wherein the style breach annotations are provided by annotators indicating locations in the training content at which style breaches are perceived.