# US 11,443,115 B2 · Machine learning from tone analysis in online customer service

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

- Number: US 11,443,115 B2
- Document ID: US11443115B2
- Status: Granted
- Kind: B2
- Application number: 16/795,401
- Filed: 2020-02-19
- Published: 2020-06-11
- Granted: 2022-09-13
- Earliest claimed: 2018-09-27
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Peifeng Yin; Zhe Liu; Anbang Xu; Taiga Nakamura
- Web page: https://patentgrep.com/patent/US11443115B2

## CPC

- G06F 40/289: Phrasal analysis, e.g. finite state techniques or chunking
- G06F 16/951: Indexing; Web crawling techniques
- G06F 40/205: Parsing
- G06F 40/30: Semantic analysis
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- 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

One embodiment provides a method that includes receiving adjusted labeled data based on emotional tone factors. Words are analyzed using a tone latent Dirichlet allocation (T-LDA) model that models tone intensity using the emotional tone factors and integrating the adjusted labeled data. Representative words are provided for each emotional tone factor based on using the T-LDA model. The representative words are obtained using the T-LDA model based on determining posterior probabilities and adjusting the posterior probabilities based on an auxiliary topic.

## Claims

### Claim 1

A method comprising:
receiving, by a processor, adjusted labeled data;
executing, by the processor, a computing factor analysis for the received adjusted labeled data to obtain and store emotional tone factors in a memory device;
executing, by the processor, a tone modified latent Dirichlet allocation (T-LDA) machine learning model that models tone intensity, wherein the T-LDA machine learning model uses the emotional tone factors and integrates the adjusted labeled data for processing words;
identifying, by the processor, representative words for each emotional tone factor by the T-LDA machine learning model, wherein the representative words are identified by the T-LDA machine learning model based on computing posterior probabilities and adjusting the posterior probabilities based on an auxiliary topic, and the T-LDA machine learning model outputs examples of tone; and
executing, by the processor, model prediction processing using the examples of tone and unlabeled data to output predicted tone for tone labeling for the unlabeled data.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
crawling a network for raw data; and
refining emotion metrics for the raw data.

### Claim 3 (depends on 2)

The method of claim 2, further comprising:
receiving, by the processor, the labeled data for the raw data using the refined emotion metrics.

### Claim 4 (depends on 3)

The method of claim 3, wherein:
executing the computing factor analysis comprises executing, by the processor, a principal components analysis (PCA) computing process on the labeled data to define the emotional tone factors; and
the emotional tone factors comprise tone metrics that are selected from the group consisting of an anxious metric, an excited metric, a frustrated metric, an impolite metric, a polite metric, a sad metric, a satisfied metric and a sympathetic metric.

### Claim 5 (depends on 4)

The method of claim 4, wherein the T-LDA machine learning model identifies corresponding descriptive words and sentences.

### Claim 6 (depends on 5)

The method of claim 5, wherein the T-LDA machine learning model uses a set of topics to identify words based on a corresponding word-distribution vector, the set of topics comprises a tone-related topic, a domain-specific topic, and the auxiliary topic, and each single word of the representative words is identified by a single topic of the set of topics.

### Claim 7 (depends on 6)

The method of claim 6, wherein the T-LDA machine learning model provides at least one of the following: providing an understanding for the emotional tone factors via the representative words, providing example sentences for each emotional tone factor or documenting topic distribution for the set of topics.

### Claim 8

A computer program product for tone analysis for online customer service, the computer program product comprising a non-statutory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
receive, by the processor, adjusted labeled data;
execute, by the processor, a computing factor analysis for the received adjusted labeled data to obtain and store emotional tone factors in a memory device;
execute, by the processor, a tone modified latent Dirichlet allocation (T-LDA) machine learning model that models tone intensity, wherein the T-LDA machine learning model uses the emotional tone factors and integrates the adjusted labeled data for processing words;
identify, by the processor, representative words for each emotional tone factor by the T-LDA machine learning model, wherein the representative words are identified by the T-LDA machine learning model based on computing posterior probabilities and adjusting the probabilities based on an auxiliary topic, and the T-LDA machine learning model outputs examples of tone; and
execute, by the processor, model prediction processing using the examples of tone and unlabeled data to output predicted tone for tone labeling for the unlabeled data.

### Claim 9 (depends on 8)

The computer program product of claim 8, wherein the program instructions executable by the processor further cause the processor to:
crawl, by the processor, a network for raw data; and
receive, by the processor, refined emotion metrics for the raw data.

### Claim 10 (depends on 9)

The computer program product of claim 9, wherein the program instructions executable by the processor further cause the processor to:
receive, by the processor, the labeled data for the raw data using the refined emotion metrics.

### Claim 11 (depends on 10)

The computer program product of claim 10, wherein:
executing the computing factor analysis comprises executing, by the processor, a principal components analysis (PCA) computing process on the labeled data to define the emotional tone factors; and
the emotional tone factors comprise tone metrics that are selected from the group consisting of an anxious metric, an excited metric, a frustrated metric, an impolite metric, a polite metric, a sad metric, a satisfied metric and a sympathetic metric.

### Claim 12 (depends on 11)

The computer program product of claim 11, wherein the T-LDA machine learning model identifies corresponding descriptive words and sentences.

### Claim 13 (depends on 12)

The computer program product of claim 12, wherein the T-LDA machine learning model uses a set of topics to identify words based on a corresponding word-distribution vector, the set of topics comprises a tone-related topic, a domain-specific topic, and the auxiliary topic, and each single word of the representative words is identified by a single topic of the set of topics.

### Claim 14 (depends on 13)

The computer program product of claim 13, wherein:
the T-LDA machine learning model provides at least one of the following: providing an understanding for the emotional tone factors via representative words, providing example sentences for each emotional tone factor or documenting topic distribution for the set of topics.

### Claim 15

An apparatus comprising:
a memory configured to store instructions; and
a processor configured to execute the instructions to:
receive adjusted labeled data;
execute a computing factor analysis for the received adjusted labeled data to obtain and store emotional tone factors in a memory device;
execute a tone latent Dirichlet allocation (T-LDA) machine learning model that models tone intensity, wherein the T-LDA machine learning model uses the emotional tone factors and integrates the adjusted labeled data for processing words;
identify representative words for each emotional tone factor by the T-LDA machine learning model, wherein the representative words are identified by the T-LDA machine learning model based on computing posterior probabilities and adjusting the probabilities based on an auxiliary topic, and the T-LDA machine learning model outputs examples of tone; and
execute model prediction processing using the examples of tone and unlabeled data to output predicted tone for tone labeling for the unlabeled data.

### Claim 16 (depends on 15)

The apparatus of claim 15, wherein the processor is further configured to execute the instructions to:
crawl a network for raw data; and
receive refined emotion metrics for the raw data.

### Claim 17 (depends on 16)

The apparatus of claim 16, wherein the processor is further configured to execute the instructions to:
receive the labeled data for the raw data using the refined emotion metrics.

### Claim 18 (depends on 17)

The apparatus of claim 17, wherein executing the computing factor analysis comprises executing, by the processor, a principal components analysis (PCA) computing process on the labeled data to define the emotional tone factors, and the emotional tone factors comprise tone metrics that are selected from the group consisting of an anxious metric, an excited metric, a frustrated metric, an impolite metric, a polite metric, a sad metric, a satisfied metric and a sympathetic metric.

### Claim 19 (depends on 18)

The apparatus of claim 18, wherein the T-LDA machine learning model identifies corresponding descriptive words and sentences, and the T-LDA machine learning model uses a set of topics to identify words based on a corresponding word-distribution vector, the set of topics comprise a tone-related topic, a domain-specific topic, and the auxiliary topic, and each single word of the representative words is identified by a single topic of the set of topics.

### Claim 20 (depends on 19)

The apparatus of claim 19, wherein:
the -LDA machine learning model provides at least one of the following: providing an understanding for the emotional tone factors via representative words, providing example sentences for each emotional tone factor or documenting topic distribution for the set of topics.