Predicting outcomes for events based on voice characteristics and content of a contact center communication

• US 9,472,188 B1
• Filed: 11/15/2013
• Issued: 10/18/2016
• Est. Priority Date: 11/15/2013
• Status: Active Grant

• Current Assignee: Alvaria Cayman
• Original Assignee: Noble Systems Corporation
• Inventors: Ouimette, Jason P., Haggerty, Christopher S.
• Primary Examiner(s): Opsasnick, Michael N
• Application Number:
  US14/080,925
• Time in Patent Office: 1,068 Days
• Field of Search: 704/244
• US Class Current: 1/1
• CPC Class Codes: G06F 40/216 using statistical methods
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 20/10 using kernel methods, e.g. support vector machines [SVM]
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06Q 10/04 Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" market predictions or forecasting for commercial activities G06Q30/0202
• CPC Class Codes: G06Q 10/10 Office automation; Time management
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 25/48 specially adapted for particular use
• CPC Class Codes: G10L 25/87 Detection of discrete points within a voice signal
• CPC Class Codes: G10L 25/90 Pitch determination of speech signals

Abstract

Various embodiments of the invention provide methods, systems, and computer-program products for predicting an outcome for an event of interest associated with a contact center communication. That is to say, various embodiments of the invention involve predicting an outcome for an event of interest associated with a party involved in a contact center communication based on characteristics and content of the communication conducted with the party by utilizing one or more classifier models.

Claims

1. A method for determining whether to conduct a follow-up communication with a party based on predicting an outcome for the follow-up communication with the party, the method comprising the steps of:

receiving a voice sample of a first communication conducted with the party in which the party is speaking about an event of interest;

extracting one or more voice characteristics from the voice sample by at least one computer processor;

applying at least a first classifier model to the one or more voice characteristics by the at least one computer processor to provide a first predicted outcome of whether the follow-up communication with the party will result in the event of interest occurring;

performing analytics on the voice sample by the at least one computer processor to detect a set of words being spoken by the party in the voice sample;

applying at least a second classifier model to the set of words by the at least one computer processor to provide a second predicted outcome of whether the follow-up communication with the party will result in the event of interest occurring;

determining by the at least one computer processor whether the first predicted outcome and the second predicted outcome match;

in response to the first predicted outcome and the second predicted outcome matching, predicting the outcome to the follow-up communication with the party as being the first predicted outcome;

in response to the first predicted outcome and the second predicted outcome not matching, predicting the outcome to the follow-up communication with the party as being (1) the first predicted outcome if the first classifier model is historically more accurate than the second classifier model at providing predicted outcomes or (2) the second predicted outcome if the second classifier model is historically more accurate than the first classifier model at providing predicted outcomes; and

conducting the follow-up communication with the party in response to predicting the outcome to the follow-up communication with the party will result in the event of interest occurring.

2. The method of claim 1, wherein the first communication is conducted between the party and a contact center agent.

3. The method of claim 1, wherein the follow-up communication has more than two possible outcomes and the first classifier model is one of a first set of classifier models comprising a classifier model for predicting each possible outcome based on the voice characteristics and the second classifier model is one of a second set of classifier models comprising a classifier model for predicting each possible outcome based on the set of words.

4. The method of claim 1, wherein the voice characteristics comprise one or more of accent shape, average pitch, contour slope, final lowering, pitch range, speech rate, stress frequency, breathiness, brilliance, loudness, pause discontinuity, and pitch discontinuity.

5. The method of claim 1, wherein the first classifier model and the second classifier model are support vector machines.

6. The method of claim 1, wherein extracting the voice characteristics from the voice sample involves:

normalizing the voice sample to eliminate speaker and recording variability; and

applying one or more digital signal processing methods to extract the voice characteristics.

7. The method of claim 1 further comprising the steps of, prior to applying the second classifier model to the set of words:

performing lemmatization on the set of words to map inflected words in the set of words to lemma forms;

eliminating common words from the set of words, the common words comprising words that solely fulfill a grammatical function as opposed to a content function that carries meaning; and

determining a frequency of occurrence of each remaining word in the set of words following the elimination of the common words, wherein at least one of the frequencies of occurrence is provided as an input to the second classifier model.

8. A non-transitory, computer-readable storage medium comprising computer-executable instructions for determining whether to conduct a follow-up communication with a party based on predicting an outcome for the follow-up communication with the party that when executed by at least one computer processor are configured to cause the at least one computer processor to:

receive a voice sample of a first communication conducted with the party in which the party is speaking about an event of interest;

extract one or more voice characteristics from the voice sample;

apply at least a first classifier model to the one or more voice characteristics to provide a first predicted outcome of whether the follow-up communication with the party will result in the event of interest occurring;

perform analytics on the voice sample to detect a set of words being spoken by the party in the voice sample;

apply at least a second classifier model to the set of words to provide a second predicted outcome of whether the follow-up communication with the party will result in the event of interest occurring;

determine whether the first predicted outcome and the second predicted outcome match;

in response to the first predicted outcome and the second predicted outcome matching, predict the outcome to the follow-up communication with the party as being the first predicted outcome;

in response to the first predicted outcome and the second predicted outcome not matching, predict the outcome to the follow-up communication with the party as being (1) the first predicted outcome if the first classifier model is historically more accurate than the second classifier model at providing predicted outcomes or (2) the second predicted outcome if the second classifier model is historically more accurate than the first classifier model at providing predicted outcomes; and

initiate having the follow-up communication conducted with the party in response to predicting the outcome to the follow-up communication with the party will result in the event of interest occurring.

9. The non-transitory, computer-readable storage medium of claim 8, wherein the first communication is conducted between the party and a contact center agent.

10. The non-transitory, computer-readable storage medium of claim 8, wherein the follow-up communication has more than two possible outcomes and the first classifier model is one of a first set of classifier models comprising a classifier model for predicting each possible outcome based on the voice characteristics and the second classifier model is one of a second set of classifier models comprising a classifier model for predicting each possible outcome based on the set of words.

11. The non-transitory, computer-readable storage medium of claim 8, wherein the voice characteristics comprise one or more of accent shape, average pitch, contour slope, final lowering, pitch range, speech rate, stress frequency, breathiness, brilliance, loudness, pause discontinuity, and pitch discontinuity.

12. The non-transitory, computer-readable storage medium of claim 8, wherein the first classifier model and the second classifier model are support vector machines.

13. The non-transitory, computer-readable storage medium of claim 8, wherein the computer-executable instructions are further configured, when executed, to cause the at least one computer processor to:

normalize the voice sample to eliminate speaker and recording variability; and

apply one or more digital signal processing methods to extract the voice characteristics.

14. The non-transitory, computer-readable storage medium of claim 8, wherein the computer-executable instructions are further configured, when executed, to cause the at least one computer processor to extract the voice characteristics from the voice sample prior to applying the second classifier model to the set of words by:

performing lemmatization on the set of words to map inflected words in the set of words to lemma forms;

eliminating common words from the set of words, the common words comprising words that solely fulfill a grammatical function as opposed to a content function that carries meaning; and

determining a frequency of occurrence of each remaining word in the set of words following the elimination of the common words, wherein at least one of the frequencies of occurrence is provided as an input to the second classifier model.

15. A system for determining whether to conduct a follow-up communication with a party based on predicting an outcome for the follow-up communication with the party, the system comprising:

At least one computer processor configured to;

receive a voice sample of a first communication conducted with the party in which the party is speaking about an event of interest;

extract one or more voice characteristics from the voice sample;

apply at least a first classifier model to the one or more voice characteristics to provide a first predicted outcome of whether the follow-up communication with the party will result in the event of interest occurring;

perform analytics on the voice sample to detect a set of words being spoken by the party in the voice sample;

apply at least a second classifier model to the set of words to provide a second predicted outcome of whether the follow-up communication with the party will result in the event of interest occurring;

determine whether the first predicted outcome and the second predicted outcome match;

in response to the first predicted outcome and the second predicted outcome matching, predict the outcome to the follow-up communication with the party as being the first predicted outcome;

in response to the first predicted outcome and the second predicted outcome not matching, predict the outcome to the follow-up communication with the party as being (1) the first predicted outcome if the first classifier model is historically more accurate than the second classifier model at providing predicted outcomes or (2) the second predicted outcome if the second classifier model is historically more accurate than the first classifier model at providing predicted outcomes; and

initiate having the follow-up communication conducted with the party in response to predicting the outcome to the follow-up communication with the party will result in the event of interest occurring.

16. The system of claim 15, wherein the first communication is conducted between the party and a contact center agent.

17. The system of claim 15, wherein the follow-up communication has more than two possible outcomes and the first classifier model is one of a first set of classifier models comprising a classifier model for predicting each possible outcome based on the voice characteristics and the second classifier model is one of a second set of classifier models comprising a classifier model for predicting each possible outcome based on the set of words.

18. The system of claim 15, wherein the voice characteristics comprise one or more of accent shape, average pitch, contour slope, final lowering, pitch range, speech rate, stress frequency, breathiness, brilliance, loudness, pause discontinuity, and pitch discontinuity.

19. The system of claim 15, wherein the first classifier model and the second classifier model are support vector machines.

20. The system of claim 15, wherein the at least one computer processor is further configured to:

normalize the voice sample to eliminate speaker and recording variability; and

apply one or more digital signal processing methods to extract the voice characteristics.

21. The system of claim 15, wherein at least one computer processor is configured to extract the voice characteristics from the voice sample prior to applying the second classifier model to the set of words by:

performing lemmatization on the set of words to map inflected words in the set of words to lemma forms;

eliminating common words from the set of words, the common words comprising words that solely fulfill a grammatical function as opposed to a content function that carries meaning; and

determining a frequency of occurrence of each remaining word in the set of words following the elimination of the common words, wherein at least one of the frequencies of occurrence is provided as an input to the second classifier model.