Systems and methods for selecting effective phrases to be presented during a conversation
A conversation may be monitored in real time using a trained machine learning model to identify a desired outcome of a conversation and generate one or more phrases for accomplishing the desired outcome. A confidence score may also be determined for one or more phrases that indicates a likelihood that the one or more phrases may help accomplish the desired outcome of the conversation. In some examples, a confidence score may be based on whether an agent, a caller, or both responded unfavorably to a similar phrase used previously in another conversation. In other examples, a confidence score corresponding to one or more phrases may be based on whether a prior conversation in which one or more similar phrases was used resulted in the desired outcome being accomplished.
1. One or more non-transitory computer-readable media storing instructions, which when executed by one or more hardware processors, cause performance of operations comprising:
training a machine learning model to identify a phrase for use in a conversation at least by:
obtaining training data sets, each training data set comprising:
a set of prior conversations;
a set of phrases used in at least one prior conversation of the set of prior conversations;
a set of prior desired outcomes corresponding to the prior conversations of the set of prior conversations;
a set of rates at which the prior desired outcomes of the prior conversations were accomplished;
training the machine learning model based on the training data sets;
monitoring a conversation in real-time to detect one or more attributes of the conversation;
identifying a set of one or more candidate phrases for accomplishing a desired outcome of the conversation;
applying the trained machine model to select a particular phrase from the set of one or more candidate phrases, the particular phrase associated with an increased likelihood of accomplishing the desired outcome of the conversation;
presenting the particular phrase for use in the conversation;
updating the training data sets by:
identifying a first start date and a first end date between which the set of prior conversations occurred;
identifying (a) a second start date between the first start date and the first end date and (b) a second end date after the first end date;
selecting an updated set of prior conversations that occurred between the second start date and the second end date; and
retraining the trained machine learning model with the updated set of prior conversations.
2. One or more non-transitory computer-readable media storing instructions, which when executed by one or more hardware processors, cause performance of operations comprising:
training a machine learning model to identify a phrase for use in a conversation at least by:
obtaining training data sets, each training data set comprising:
a set of prior conversations;
a set of phrases used in at least one prior conversation of the set of prior conversations;
a set of prior desired outcomes corresponding to the prior conversations of the set of prior conversations;
a set of rates at which the prior desired outcomes of the prior conversations were accomplished;
training the machine learning model based on the training data sets;
monitoring a conversation in real-time to detect one or more attributes of the conversation;
identifying a set of one or more candidate phrases for accomplishing a desired outcome of the conversation;
applying the trained machine model to select a particular phrase from the set of one or more candidate phrases, the particular phrase associated with an increased likelihood of accomplishing the desired outcome of the conversation;
presenting the particular phrase for use in the conversation;
monitoring the conversation after providing the particular phrase, the monitoring identifying at least one additional attribute of the conversation;
responsive to the at least one additional attribute, re-applying the trained machine learning model;
based on the re-application of the trained machine learning model, selecting a second phrase from the set of one or more candidate phrases to be provided for use in the conversation; and
wherein the particular phrase is de-selected and replaced by the second phrase, the second phrase being provided for use in the conversation instead of the particular phrase.
3. One or more non-transitory computer-readable media storing instructions, which when executed by one or more hardware processors, cause performance of operations comprising:
training a machine learning model to identify a phrase for use in a conversation at least by:
obtaining training data sets, each training data set comprising:
a set of prior conversations;
a set of phrases used in at least one prior conversation of the set of prior conversations;
a set of prior desired outcomes corresponding to the prior conversations of the set of prior conversations;
a set of rates at which the prior desired outcomes of the prior conversations were accomplished;
training the machine learning model based on the training data sets;
monitoring a conversation in real-time to detect one or more attributes of the conversation;
identifying a set of one or more candidate phrases for accomplishing a desired outcome of the conversation;
applying the trained machine model to select a particular phrase from the set of one or more candidate phrases, the particular phrase associated with an increased likelihood of accomplishing the desired outcome of the conversation;
presenting the particular phrase for use in the conversation;
wherein the set of phrases used in the at least one prior conversation of the set of prior conversations in the training data sets are associated with:
at least one prior desired outcome of the set of prior desired outcomes;
a corresponding label indicating whether the prior desired outcome was accomplished or not accomplished;
wherein the applying operation comprises refraining from selecting a phrase associated with the label indicating the prior desired outcome was not accomplished as the particular phrase;
the set of phrases used in the at least one prior conversation of the set of prior conversations are associated with a label indicating whether a response to the phrases of the set of phrases was an adverse response; and
the applying operation comprises refraining from selecting a phrase associated with the adverse response as the particular phrase.
4. The non-transitory computer-readable media of claim 1 , wherein:
each training data set further comprises a set of characteristics associated with prior participants in the one or more corresponding prior conversations; and
applying the trained machine learning model further comprises selecting the particular phrase based, at least in part, on a set of target characteristics associated with a present participant in the conversation.
5. The non-transitory computer-readable media of claim 4 , wherein the characteristics of the set of characteristics associated with the prior participants in one or more corresponding prior conversations comprise one or more of demographic information, a browsing history, and a product engagement history.
6. The non-transitory computer-readable media of claim 1 , wherein presenting the particular phrase for use in the conversation comprises presenting the particular phrase in a text transmission interface.
7. The non-transitory computer-readable media of claim 1 , further comprising identifying the desired outcome of the conversation based on the one or more attributes of the conversation.
8. A method comprising:
training a machine learning model to identify a phrase for use in a conversation at least by:
obtaining training data sets, each training data set comprising:
a set of prior conversations;
a set of phrases used in at least one prior conversation of the set of prior conversations;
a set of prior desired outcomes corresponding to the prior conversations of the set of prior conversations;
a set of rates at which the prior desired outcomes of the prior conversations were accomplished;
training the machine learning model based on the training data sets;
monitoring a conversation in real-time to detect one or more attributes of the conversation;
identifying a set of one or more candidate phrases for accomplishing a desired outcome of the conversation;
applying the trained machine model to select a particular phrase from the set of one or more candidate phrases, the particular phrase associated with an increased likelihood of accomplishing the desired outcome of the conversation;
presenting the particular phrase for use in the conversation;
updating the training data sets by:
identifying a first start date and a first end date between which the set of prior conversations occurred;
identifying (a) a second start date between the first start date and the first end date and (b) a second end date after the first end date;
selecting an updated set of prior conversations that occurred between the second start date and the second end date; and
retraining the trained machine learning model with the updated set of prior conversations.
9. A method comprising:
training a machine learning model to identify a phrase for use in a conversation at least by:
obtaining training data sets, each training data set comprising:
a set of prior conversations;
a set of phrases used in at least one prior conversation of the set of prior conversations;
a set of prior desired outcomes corresponding to the prior conversations of the set of prior conversations;
a set of rates at which the prior desired outcomes of the prior conversations were accomplished;
training the machine learning model based on the training data sets;
monitoring a conversation in real-time to detect one or more attributes of the conversation;
identifying a set of one or more candidate phrases for accomplishing a desired outcome of the conversation;
applying the trained machine model to select a particular phrase from the set of one or more candidate phrases, the particular phrase associated with an increased likelihood of accomplishing the desired outcome of the conversation;
presenting the particular phrase for use in the conversation;
monitoring the conversation after providing the particular phrase, the monitoring identifying at least one additional attribute of the conversation;
responsive to the at least one additional attribute, re-applying the trained machine learning model; and
based on the re-application of the trained machine learning model, selecting a second phrase from the set of one or more candidate phrases to be provided for use in the conversation; and
wherein the particular phrase is de-selected and replaced by the second phrase, the second phrase being provided for use in the conversation instead of the particular phrase.
10. A method comprising:
training a machine learning model to identify a phrase for use in a conversation at least by:
obtaining training data sets, each training data set comprising:
a set of prior conversations;
a set of phrases used in at least one prior conversation of the set of prior conversations;
a set of prior desired outcomes corresponding to the prior conversations of the set of prior conversations;
a set of rates at which the prior desired outcomes of the prior conversations were accomplished;
training the machine learning model based on the training data sets;
monitoring a conversation in real-time to detect one or more attributes of the conversation;
identifying a set of one or more candidate phrases for accomplishing a desired outcome of the conversation;
applying the trained machine model to select a particular phrase from the set of one or more candidate phrases, the particular phrase associated with an increased likelihood of accomplishing the desired outcome of the conversation;
presenting the particular phrase for use in the conversation;
wherein the set of phrases used in the at least one prior conversation of the set of prior conversations in the training data sets are associated with:
at least one prior desired outcome of the set of prior desired outcomes;
a corresponding label indicating whether the prior desired outcome was accomplished or not accomplished; and
wherein the applying operation comprises refraining from selecting a phrase associated with the label indicating the prior desired outcome was not accomplished as the particular phrase;
the set of phrases used in the at least one prior conversation of the set of prior conversations are associated with a label indicating whether a response to the phrases of the set of phrases was an adverse response; and
the applying operation comprises refraining from selecting a phrase associated with the adverse response as the particular phrase.
11. The method of claim 8 , wherein:
each training data set further comprises a set of characteristics associated with prior participants in the one or more corresponding prior conversations; and
applying the trained machine learning model further comprises selecting the particular phrase based, at least in part, on a set of target characteristics associated with a present participant in the conversation.
12. The method of claim 11 , wherein the characteristics of the set of characteristics associated with the prior participants in one or more corresponding prior conversations comprise one or more of demographic information, a browsing history, and a product engagement history.
13. The method of claim 8 , wherein presenting the particular phrase for use in the conversation comprises presenting the particular phrase in a text transmission interface.
14. The method of claim 8 , further comprising identifying the desired outcome of the conversation based on the one or more attributes of the conversation.