System and method for automated root cause investigation

1. A computer-implemented method of analyzing recorded interactions, the method comprising, using one or more computer processors:

selecting a reference phrase category based on input from a user via a computer user interface;

associating by a processor of the one or more processors each of a set of recorded interactions with at least one phrase category included in a set of phrase categories stored in a database, wherein a recorded interaction is associated with a phrase category if a text of the recorded interaction includes at least one phrase included in the phrase category and wherein each phrase category includes one or more phrases, each recorded interaction stored in the database and associated with text produced by applying a speech-to-text process to an audio recording or a video recording of the recorded interaction;

receiving a selection of a first time interval, and from a user via a computer user interface a selection of a second time interval, the first time interval preceding the second time interval;

a. calculating, by a processor of the one or more processors, for each of at least some of the phrase categories in the set of phrase categories a correlation differential by:

calculating, for the first time interval, a first correlation value for the phrase category and the reference phrase category, the first correlation value representing the correlation of a trend of the phrase category with a trend of the reference phrase category, wherein a trend of a phrase category is defined by a set of values representing the number of recorded interactions, in a respective set of time units, that are associated with the phrase category;

calculating, for the second time interval, a second correlation value for the phrase category and the reference phrase category, and

calculating the correlation differential of the phrase category by subtracting the second correlation value from the first correlation value;

b. selecting, by a processor of the one or more processors, a candidate phrase category by identifying the phrase category with the highest correlation differential and denoting the correlation differential of the candidate phrase category as the candidate correlation differential;

adding, by a processor of the one or more processors, at least one of the phrases included in the candidate phrase category to a root cause list of phrases;

using a search engine executed by a processor of the one or more processors, determining a number of interactions including a phrase from the root cause list of phrases; and

displaying to the user via the computer user interface the root cause list of phrases and the number of interactions.

2. The method of claim 1 , comprising selecting the candidate phrase category by identifying the phrase category with the highest correlation differential value and whose trend changed, in the second time interval with respect to the first time interval, in the same direction of a change of the trend of the reference phrase category.

3. The method of claim 1 , comprising:

c. for each phrase category with a negative correlation differential:

removing the phrase category from the set of phrase categories by removing recorded interactions associated with the phrase category from the set of recorded interactions;

calculating an updated candidate correlation value for the candidate phrase category and the reference phrase category, and

if the difference between the candidate correlation value and the updated candidate correlation value is below a threshold then returning the phrase category to the set of phrase categories by returning the recorded interactions that were removed at step (a) to the set of recorded interactions; and

if the difference between the candidate correlation value and the updated candidate correlation value is above a threshold then adding at least one of the phrases included in the candidate phrase category to the root cause list of phrases.

4. The method of claim 3 , comprising:

if the correlation differential value of the reference phrase category and the candidate phrase category is above a threshold then producing a combined phrase category by combining the candidate phrase category and the reference phrase category, denoting the combined phrase category as the reference phrase category and repeating steps b and c.

5. The method of claim 4 , comprising:

selecting a phrase from the combined phrase category;

generating a phrase trend for the phrase based on the number of recorded interactions, in the set of recorded interactions, that include the phrase, per time unit, during the second time interval;

generating a combined phrase category trend for the combined phrase category;

determining a correlation level by relating the phrase trend to the combined phrase category trend; and

if the correlation level is greater than a threshold level then including the phrase in the root cause list of phrases and providing the list to a user.

6. The method of claim 4 , comprising:

for each phrase category in the combined phrase category:

for each phrase in the category:

if a correlation of the phrase trend of the phrase with a trend of the phrase category is greater than the correlation of the phrase trend with a trend of the combined phrase category then removing the phrase category from the combined phrase category.

7. The method of claim 5 , comprising:

for at least one phrase included in in the root cause list of phrases:

calculating a phrase correlation differential based on a trend of the phrase and a trend of the combined phrase category; and

if the correlation differential is less than a threshold level then removing the phrase from the root cause list of phrases.

8. The method of claim 5 , comprising iteratively selecting all of the phrases in the combined phrase category.

9. The method of claim 1 , comprising:

providing, with respect to the input related to the reference phrase category, at least one of: a list of interactions, a number of interactions, a list of related categories and a list of related phrases.

10. The method of claim 1 , wherein a recorded interaction is associated with a phrase category if at least one phrase included in the phrase category is included in the recorded interaction.

11. The method of claim 1 , wherein a recorded interaction is associated with a phrase category based on metadata related to the recorded interaction.

12. A computer-implemented method of automated identification of a cause of a problem, the method comprising, using one or more computer processors:

selecting a reference phrase category based on input from a user via a computer user interface;

associating by a processor of the one or more processors each of a set of recorded interactions with at least one phrase category stored in a database, wherein the at least one phrase category includes one or more phrases and wherein a recorded interaction is associated with a phrase category if the a text of the recorded interaction includes at least one phrase included in the phrase category, each recorded interaction stored in the database and associated with text produced by applying a speech-to-text process to an audio recording or a video recording of the recorded interaction;

selecting a base phrase category;

calculating by a processor of the one or more processors, for each of at least some phrase categories in a set of phrase categories a correlation differential by:

calculating, for a first time interval, a first correlation level for the phrase category and the base phrase category, the first correlation level quantifying the correlation level of a trend of the phrase category with a trend of the reference phrase category, wherein a trend wherein a trend includes values over time;

calculating, for a second time interval, a second correlation level for the phrase category and the reference phrase category, and

calculating the correlation differential of the phrase category by comparing the first correlation level and the second correlation level;

selecting by a processor of the one or more processors a candidate phrase category by identifying, in the set of categories, the phrase category with the largest correlation differential and denoting the correlation differential of the candidate phrase category as the candidate correlation differential;

for each phrase category for which a negative correlation differential was calculated:

removing by a processor of the one or more processors the phrase category from the set of phrase categories;

calculating an updated candidate correlation level for the candidate phrase category and the reference phrase category, and

if the difference between the candidate correlation level and the updated candidate correlation level is below a threshold then returning the phrase category to the set of phrase categories; and

if the difference between the candidate correlation level and the updated candidate correlation level is above a threshold then adding at least one of the phrases included in the candidate phrase category to a root cause list of phrases;

using a search engine executed by a processor of the one or more processors, determining a number of interactions including a phrase from the root cause list of phrases; and

displaying to the user via the computer user interface the root cause list of phrases and the number of interactions.

13. A system comprising:

a memory; and

a controller configured to:

define a set of phrase categories by including one or more phrases in each of the phrase categories, the phrase categories stored in a database;

associate each of a set of recorded interactions with at least one phrase category in the set of phrase categories, wherein a recorded interaction is associated with a phrase category if the recorded interaction includes at least one phrase included in the phrase category, each recorded interaction stored in the database and associated with text produced by applying a speech-to-text process to an audio recording or a video recording of the recorded interaction;

receive a selection of a first time interval and from a user via a computer user interface a selection of a second time interval, the first time interval preceding the second time interval;

select a reference phrase category based on input from a user via a computer user interface;

a. calculate, for each of at least some of the phrase categories in the set of phrase categories a correlation differential by:

calculating, for the first time interval, a first correlation value for the phrase category and the reference phrase category, the first correlation value representing the correlation of a trend of the phrase category with a trend of the reference phrase category, wherein a trend wherein a trend includes values over time;

calculating, for the second time interval, a second correlation value for the phrase category and the reference phrase category, and

calculating the correlation differential of the phrase category by subtracting the second correlation value from the first correlation value;

b. select a candidate phrase category by identifying the phrase category with the highest correlation differential and denoting the correlation differential of the candidate phrase category as the candidate correlation differential; and

add at least one of the phrases included in the candidate phrase category to a root cause list of phrases;

use a search engine to determine a number of interactions including a phrase from the root cause list of phrases; and

display to the user via the computer user interface the root cause list of phrases and the number of interactions.

14. The system of claim 13 , wherein the controller is further configured to select the candidate phrase category by identifying the phrase category with the highest correlation differential value and whose trend changed, in the second time interval with respect to the first time interval, in the same direction of a change of the trend of the reference phrase category.

15. The system of claim 13 , wherein the controller is further configured to:

c. for each phrase category with a negative correlation differential:

remove the phrase category from the set of phrase categories by removing recorded interactions associated with the phrase category from the set of recorded interactions;

calculate an updated candidate correlation value for the candidate phrase category and the reference phrase category, and

if the difference between the candidate correlation value and the updated candidate correlation value is below a threshold then return the phrase category to the set of phrase categories by returning the recorded interactions that were removed at step (a) to the set of recorded interactions; and

if the difference between the candidate correlation value and the updated candidate correlation value is above a threshold then add at least one of the phrases included in the candidate phrase category to the root cause list of phrases.

16. The system of claim 15 , wherein the controller is further configured to, if the correlation differential value of the reference phrase category and the candidate phrase category is above a threshold then produce a combined phrase category by combining the candidate phrase category and the reference phrase category, denoting the combined phrase category as the reference phrase category and repeating steps b and c.

17. The system of claim 14 , wherein the controller is further configured to:

select a phrase from the combined phrase category;

generate a phrase trend for the phrase based on the number of recorded interactions, in the set of recorded interactions, that include the phrase, per time unit, during the second time interval;

generate a combined phrase category trend for the combined phrase category;

determine a correlation level by relating the phrase trend to the combined phrase category trend; and

if the correlation level is greater than a threshold level then include the phrase in a root cause list of phrases and provide the list to a user.

18. The system of claim 16 , wherein the controller is further configured to:

for each phrase category in the combined phrase category:

for each phrase in the category:

if a correlation of the phrase trend of the phrase with a trend of the phrase category is greater than the correlation of the phrase trend with a trend of the combined phrase category then remove the phrase category from the combined phrase category.

19. The system of claim 15 , wherein the controller is further configured to:

for at least one phrase included in in the root cause list of phrases:

calculate a phrase correlation differential based on a trend of the phrase and a trend of the combined phrase category; and

if the correlation differential is less than a threshold level then remove the phrase from the root cause list of phrases.

20. The system of claim 16 , wherein the controller is further configured to iteratively select all of the phrases in the combined phrase category.

21. The system of claim 13 , wherein the controller is further configured to:

select the reference phrase category based on input from a user; and

provide, with respect to the input, at least one of: a list of interactions, a number of interactions, a list of related categories and a list of related phrases.