Machine learning from tone analysis in online customer service

1. A method comprising:

crawling a network for raw data;

refining emotion metrics for the raw data;

receiving labels for the raw data using refined emotion metrics;

performing factor analysis for labeled data to obtain emotional tone factors;

receiving adjusted labeled data based on the emotional tone factors;

analyzing words using a tone latent Dirichlet allocation (T-LDA) model that models tone intensity using the emotional tone factors and integrating the adjusted labeled data; and

provide representative words for each emotional tone factor based on using the tone model, wherein the representative words are obtained using the T-LDA model based on determining posterior probabilities and adjusting the probabilities based on an auxiliary topic.

2. The method of claim 1 , wherein performing the factor analysis comprises performing, by a processor, a principal components analysis (PCA) on the labeled data to define the emotional tone factors.

3. The method of claim 1 , wherein the emotional tone factors are selected from the group consisting of anxious, excited, frustrated, impolite, polite, sad, satisfied and sympathetic.

4. The method of claim 3 , wherein the T-LDA model determines corresponding descriptive words and sentences.

5. The method of claim 4 , wherein the T-LDA model uses a set of topics to generate words based on a corresponding word-distribution vector.

6. The method of claim 5 , wherein the set of topics comprise a tone-related topic, a domain-specific topic, and the auxiliary topic.

7. The method of claim 6 , wherein:

the tone 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.

8. A computer program product for tone analysis for online customer service, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

crawl, by the processor, a network for raw data;

receive, by the processor, refined emotion metrics for the raw data;

receive, by the processor, labels for the raw data using refined emotion metrics;

perform, by the processor, factor analysis for labeled data to obtain emotional tone factors;

receive, by the processor, adjusted labeled data based on the emotional tone factors;

analyze, by the processor, words using a tone latent Dirichlet allocation (T-LDA) model that models tone intensity using the emotional tone factors and integrating the adjusted labeled data; and

provide, by the processor, representative words for each emotional tone factor based on using the tone model, wherein the representative words are obtained using the T-LDA model based on determining posterior probabilities and adjusting the probabilities based on an auxiliary topic.

9. The computer program product of claim 8 , wherein performing the factor analysis comprises performing, by a processor, a principal components analysis (PCA) on the labeled data to define the emotional tone factors.

10. The computer program product of claim 8 , wherein the emotional tone factors are selected from the group consisting of anxious, excited, frustrated, impolite, polite, sad, satisfied and sympathetic.

11. The computer program product of claim 10 , wherein the T-LDA model determines corresponding descriptive words and sentences.

12. The computer program product of claim 11 , wherein the T-LDA model uses a set of topics to generate words based on a corresponding word-distribution vector.

13. The computer program product of claim 12 , wherein the set of topics comprise a tone-related topic, a domain-specific topic, and the auxiliary topic.

14. The computer program product of claim 13 , wherein:

the tone 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.

15. An apparatus comprising:

a memory configured to store instructions; and

a processor configured to execute the instructions to:

crawl a network for raw data;

receive refined emotion metrics for the raw data;

receive labels for the raw data using refined emotion metrics;

perform factor analysis for labeled data to obtain emotional tone factors;

receive adjusted labeled data based on the emotional tone factors;

analyze words using a tone latent Dirichlet allocation (T-LDA) model that models tone intensity using the emotional tone factors and integrating the adjusted labeled data; and

provide representative words for each emotional tone factor based on using the tone model, wherein the representative words are obtained using the T-LDA model based on determining posterior probabilities and adjusting the probabilities based on an auxiliary topic.

16. The apparatus of claim 15 , wherein performing the factor analysis comprises performing, by a processor, a principal components analysis (PCA) on the labeled data to define the emotional tone factors.

17. The apparatus of claim 15 , wherein the emotional tone factors are selected from the group consisting of anxious, excited, frustrated, impolite, polite, sad, satisfied and sympathetic.

18. The apparatus of claim 17 , wherein the T-LDA model determines corresponding descriptive words and sentences.

19. The apparatus of claim 18 , wherein the T-LDA model uses a set of topics to generate words based on a corresponding word-distribution vector, and the set of topics comprise a tone-related topic, a domain-specific topic, and the auxiliary topic.

20. The apparatus of claim 19 , wherein:

the tone 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.