Automated support center call topic identification

1 . A method comprising:

generating, by a computing system based on a first plurality of digitized voice signal recordings, a plurality of first datasets that correspond to ones of the first plurality of digitized voice signal recordings, each digitized voice signal recording in the first plurality of digitized voice signal recordings corresponding to a support call of a first plurality of support calls between a technician and a support representative providing support to the technician, each first dataset comprising one or more textual documents that comprise words spoken during the corresponding support call between the technician and the support representative;

training, by the computing system using the plurality of first datasets, an unsupervised learning model operative to identify clusters of first datasets from the plurality of first datasets based on similarity to a corresponding topic of a plurality of first topics, each first topic of the plurality of first topics comprising a first set of topic terms;

obtaining, by the computing system from the unsupervised learning model, the plurality of first topics, each first topic of the plurality of first topics comprising information indicative of a corresponding first set of topic terms and a respective cluster of first datasets;

assigning, by the computing system, to a respective first topic of the plurality of first topics, a textual description based on the corresponding first set of topic terms, wherein the textual description comprises a category and a subcategory of a plurality of subcategories corresponding to the category, wherein the category and the subcategory are descriptive of the respective first topic and the respective cluster of first datasets; and

outputting information that corresponds to at least one first topic of the plurality of first topics.

2 . The method of claim 1 further comprising:

generating, based on a second plurality of digitized voice signal recordings, a plurality of second datasets that correspond to ones of the second plurality of digitized voice signal recordings, each digitized voice signal recording in the second plurality of digitized voice signal recordings corresponding to a support call of a second plurality of support calls between a technician and a support representative providing support to the technician, each second dataset comprising one or more textual documents that comprise words spoken during the corresponding support call between the technician and the support representative;

training, by the computing system using the plurality of second datasets, an unsupervised learning model operative to identify clusters of second datasets from the plurality of second datasets based on similarity to a corresponding topic of a plurality of second topics, each second topic of the plurality of second topics comprising a second set of topic terms;

determining, by the computing system using a similarity function, that a similarity value between a particular second topic and a particular first topic is above a predetermined threshold; and

assigning, to each second dataset in a cluster of second datasets, a textual description that corresponds to the particular first topic.

3 . The method of claim 2 wherein determining, by the computing system using the similarity function, that the similarity value between the particular second topic and the particular first topic is above the predetermined threshold comprises:

determining, by the computing system, a similarity value between the particular second topic and each first topic of the plurality of first topics; and

determining that the particular first topic is most similar to the particular second topic.

4 . The method of claim 2 further comprising:

determining, by the computing system, a similarity value between an another second topic and each first topic of the plurality of first topics to generate a plurality of similarity values;

determining that none of the plurality of similarity values are above the predetermined threshold; and

outputting information indicating that the another second topic is not similar to any other first topic.

5 . The method of claim 2 further comprising:

determining, for a set of second topics, a corresponding set of first topics of the plurality of first topics, each second topic in the set of second topics having a similarity value with a first topic in the corresponding set of first topics above the predetermined threshold; and

assigning, to each second dataset corresponding to a second topic in the set of second topics, a textual description assigned to the corresponding first topic.

6 . The method of claim 5 further comprising:

outputting, by the computing system, information that identifies each respective textual description and a quantity of second datasets to which the respective textual description has been assigned.

7 . A computing system comprising:

one or more computing devices operable to:

generate, based on a first plurality of digitized voice signal recordings, a plurality of first datasets that correspond to ones of the first plurality of digitized voice signal recordings, each digitized voice signal recording in the first plurality of digitized voice signal recordings corresponding to a support call of a first plurality of support calls between a technician and a support representative providing support to the technician, each first dataset comprising one or more textual documents that comprise words spoken during the corresponding support call between the technician and the support representative;

train, using the plurality of first datasets, an unsupervised learning model operative to identify clusters of first datasets from the plurality of first datasets based on similarity to a corresponding topic of a plurality of first topics, each first topic of the plurality of first topics comprising a first set of topic terms;

obtain, from the unsupervised learning model, the plurality of first topics, each first topic of the plurality of first topics comprising information indicative of a corresponding first set of topic terms and a respective cluster of first datasets;

assign, to a respective first topic of the plurality of first topics, a textual description based on the corresponding first set of topic terms, wherein the textual description comprises a category and a subcategory of a plurality of subcategories corresponding to the category, wherein the category and the subcategory are descriptive of the respective first topic and the respective cluster of first datasets; and

output information that corresponds to at least one first topic of the plurality of first topics.

8 . The computing system of claim 7 wherein the one or more computing devices are further operable to:

generate, based on a second plurality of digitized voice signal recordings, a plurality of second datasets that correspond to ones of the second plurality of digitized voice signal recordings, each digitized voice signal recording in the second plurality of digitized voice signal recordings corresponding to a support call of a second plurality of support calls between a technician and a support representative providing support to the technician, each second dataset comprising one or more textual documents that comprise words spoken during the corresponding support call between the technician and the support representative;

train, using the plurality of second datasets, an unsupervised learning model operative to identify clusters of second datasets from the plurality of second datasets based on similarity to a corresponding topic of a plurality of second topics, each second topic of the plurality of second topics comprising a second set of topic terms;

determine, using a similarity function, that a similarity value between a particular second topic and a particular first topic is above a predetermined threshold; and

assign, to each second dataset in a cluster of second datasets, the textual description that corresponds to the particular first topic.

9 . The computing system of claim 8 wherein, to determine, by the one or more computing devices using the similarity function, that the similarity value between the particular second topic and the particular first topic is above the predetermined threshold, the one or more computing devices are further operable to:

determine a similarity value between the particular second topic and each first topic of the plurality of first topics; and

determine that the particular first topic is most similar to the particular second topic.

10 . The computing system of claim 8 wherein the one or more computing devices are further operable to:

determine a similarity value between an another second topic and each first topic of the plurality of first topics to generate a plurality of similarity values;

determine that none of the plurality of similarity values are above the predetermined threshold; and

output information indicating that the another second topic is not similar to any other first topic.

11 . The computing system of claim 8 wherein the one or more computing devices are further operable to:

determine, for a set of second topics, a corresponding set of first topics of the plurality of first topics, each second topic in the set of second topics having a similarity value with a first topic in the corresponding set of first topics above the predetermined threshold; and

assign, to each second dataset corresponding to a second topic in the set of second topics, a textual description assigned to the corresponding first topic.

12 . A non-transitory computer-readable storage medium that includes executable instructions operable to cause one or more computing devices to:

generate, based on a first plurality of digitized voice signal recordings, a plurality of first datasets that correspond to ones of the first plurality of digitized voice signal recordings, each digitized voice signal recording in the first plurality of digitized voice signal recordings corresponding to a support call of a first plurality of support calls between a technician and a support representative providing support to the technician, each first dataset comprising one or more textual documents that comprise words spoken during the corresponding support call between the technician and the support representative;

train, using the plurality of first datasets, an unsupervised learning model operative to identify clusters of first datasets from the plurality of first datasets based on similarity to a corresponding topic of a plurality of first topics, each first topic of the plurality of first topics comprising a first set of topic terms;

obtain, from the unsupervised learning model, the plurality of first topics, each first topic of the plurality of first topics comprising information indicative of a corresponding first set of topic terms and a respective cluster of first datasets;

assign, to a respective first topic of the plurality of first topics, a textual description based on the corresponding first set of topic terms, wherein the textual description comprises a category and a subcategory of a plurality of subcategories corresponding to the category, wherein the category and the subcategory are descriptive of the respective first topic and the respective cluster of first datasets; and

output information that corresponds to at least one first topic of the plurality of first topics.

13 . The non-transitory computer-readable storage medium of claim 12 wherein the instructions further cause the one or more computing devices to:

generate, based on a second plurality of digitized voice signal recordings, a plurality of second datasets that correspond to ones of the second plurality of digitized voice signal recordings, each digitized voice signal recording in the second plurality of digitized voice signal recordings corresponding to a support call of a second plurality of support calls between a technician and a support representative providing support to the technician, each second dataset comprising one or more textual documents that comprise words spoken during the corresponding support call between the technician and the support representative;

train, using the plurality of second datasets, an unsupervised learning model operative to identify clusters of second datasets from the plurality of second datasets based on similarity to a corresponding topic of a plurality of second topics, each second topic of the plurality of second topics comprising a second set of topic terms;

determine, using a similarity function, that a similarity value between a particular second topic and a particular first topic is above a predetermined threshold; and

assign, to each second dataset in a cluster of second datasets, the textual description that corresponds to the particular first topic.

14 . The non-transitory computer-readable storage medium of claim 13 wherein, to determine, using the similarity function, that the similarity value between the particular second topic and the particular first topic is above the predetermined threshold, the instructions further cause the one or more computing devices to:

determine a similarity value between the particular second topic and each first topic of the plurality of first topics; and

determine that the particular first topic is most similar to the particular second topic.

15 . The non-transitory computer-readable storage medium of claim 13 wherein the instructions further cause the one or more computing devices to:

determine a similarity value between an another second topic and each first topic of the plurality of first topics to generate a plurality of similarity values;

determine that none of the plurality of similarity values are above the predetermined threshold; and

output information indicating that the another second topic is not similar to any other first topic.

16 . The non-transitory computer-readable storage medium of claim 13 wherein the instructions further cause the one or more computing devices to:

determine, for a set of second topics a corresponding set of first topics of the plurality of first topics, each second topic in the set of second topics having a similarity value with a first topic in the corresponding set of first topics above the predetermined threshold; and

assign, to each second dataset corresponding to a second topic in the set of second topics, a textual description assigned to the corresponding first topic.