Modeling voice calls to improve an outcome of a call between a representative and a customer
A call-modeling system models calls in real-time, with the goal of helping users, e.g., a sales representative and/or their managers, improve and/or guide the outcome of the calls. The call-modeling system generates real-time probabilities for possible outcomes of the conversation, as well as highlight specific on-call patterns, which may be either conducive or detrimental to a desired conversation outcome. The generated probabilities and highlighted patterns may be used by the sales representatives and/or their managers to either increase the probability of a desired outcome and/or optimize for call duration with a specific outcome.
1. A computer-implemented method, comprising:
receiving real-time call data of a call between a user and a participant of multiple participants;
extracting a set of features from the real-time call data, wherein the set of features includes characteristics of the user and characteristics associated with a conversation of the call;
analyzing the set of features using multiple classifiers to determine multiple outcomes of the call, wherein the multiple classifiers are generated based on multiple features extracted from a set of historical call data; and
generating, based on the analyzing of the set of features, an on-call guidance to increase a probability of a specified outcome of the multiple outcomes.
2. The computer-implemented method of claim 1 , wherein generating the on-call guidance includes generating a notification that indicates the participant to adopt, desist, or persist with a conversation characteristic to increase or decrease the probability of the specified outcome.
3. The computer-implemented method of claim 2 , wherein the conversation characteristic includes a characteristic associated with the participant.
4. The computer-implemented method of claim 2 , wherein the conversation characteristic includes a characteristic associated with the user-participant pair.
5. The computer-implemented method of claim 2 , wherein generating the on-call guidance includes:
determining that the set of features changed as the call progressed, and
adjusting the on-call guidance to indicate the participant to adopt a specified conversation characteristic that is different from the conversation characteristic.
6. The computer-implemented method of claim 1 , wherein generating the on-call guidance includes:
analyzing the set of features with a subset of the multiple classifiers that include one or more features with a prediction power exceeding a specified threshold, and
presenting values of the one or more features as the on-call guidance.
7. The computer-implemented method of claim 1 , wherein extracting the set of features includes generating the set of features using at least one of natural language processing, artificial intelligence, or machine learning techniques.
8. The computer-implemented method of claim 1 , wherein analyzing the set of features using the multiple classifiers includes:
generating the multiple classifiers based on multiple features extracted from a set of calls between multiple users and the multiple participants, wherein each of the multiple classifiers indicates an outcome of a corresponding conversation between a specified user of the multiple users and a specified participant of the multiple participants.
9. The computer-implemented method of claim 8 , wherein generating the multiple classifiers using the set of calls includes analyzing audio recordings of conversations between the multiple users and the multiple participants.
10. The computer-implemented method of claim 1 , wherein extracting the set of features includes:
generating features that include a transcription, a vocabulary and a language model of the conversation as a first output.
11. The computer-implemented method of claim 10 , wherein extracting the set of features includes:
generating, using the first output, features that include semantic information from the conversation.
12. The computer-implemented method of claim 1 , wherein extracting the set of features includes:
generating features that include data regarding conversation flow.
13. The computer-implemented method of claim 1 , wherein extracting the set of features includes:
generating a speaker engagement metric that includes information regarding a degree of engagement of the user in the conversation.
14. The computer-implemented method of claim 1 , wherein extracting the set of features from the real-time call data includes:
automatically generating multiple tags for the call, wherein each of the multiple tags indicates a key moment of the call.
15. The computer-implemented method of claim 14 , wherein each of the multiple tags includes at least one of information regarding a moment to which the tag corresponds, a time interval at which the moment occurred in the call, a duration for which the moment lasted, or information regarding the participant of the call.
16. The computer-implemented method of claim 15 further comprising:
sharing the multiple tags with the participant during the call.
17. The computer-implemented method of claim 1 , wherein receiving the real-time call data of the call includes receiving the real-time call data of a video call between the user and the participant.
18. The computer-implemented method of claim 1 , wherein receiving the real-time call data of the call includes receiving the real-time call data of an online meeting between the user and the participant.
19. The computer-implemented method of claim 1 , wherein receiving the real-time call data of the call includes receiving the real-time call data of a virtual reality-based conversation between the user and the participant.
20. The computer-implemented method of claim 1 , wherein receiving the real-time call data of the call includes receiving the real-time call data of an augmented reality based conversation between the user and the participant.
21. The computer-implemented method of claim 1 , wherein receiving the real-time call data of the call includes receiving the real-time call data of an e-mail conversation between the user and the participant.
22. A non-transitory computer-readable storage medium storing computer-readable instructions, comprising:
instructions for receiving real-time call data of a call between a user and a participant of multiple participants;
instructions for extracting a set of features from the real-time call data, wherein the set of features includes characteristics of the user and characteristics associated with a conversation of the call;
instructions for analyzing the set of features using multiple classifiers, wherein the multiple classifiers are generated based on multiple features extracted from a set of historical call data, wherein each of the multiple classifiers indicates one of multiple outcomes of a conversation; and
instructions for generating, based on the analyzing of the set of features, an on-call guidance which, when adopted or persisted by the participant, increases a probability of a specified outcome of the multiple outcomes.
23. The non-transitory computer-readable storage medium of claim 22 , wherein the instructions for generating the multiple classifiers include:
instructions for generating a notification that indicates the participant to adopt, desist, or persist with a conversation characteristic to increase or decrease the probability of the specified outcome.
24. The non-transitory computer-readable storage medium of claim 23 , wherein the instructions for generating the multiple classifiers include:
instructions for determining that the set of features changed as the call progressed, and
instructions for adjusting the on-call guidance to indicate the participant to adopt a specified conversation characteristic that is different from the conversation characteristic.
25. The non-transitory computer-readable storage medium of claim 22 , wherein the instructions for generating the on-call guidance includes:
instructions for analyzing the set of features with a subset of the multiple classifiers that include one or more features with a prediction power exceeding a specified threshold, and
instructions for presenting values of the one or more features as the on-call guidance.
26. A system, comprising:
a first component that is configured to receive real-time call data of a call between a user and a participant of multiple participants;
a second component that is configured to extract a set of features from the real-time call data, wherein the set of features includes characteristics of the user and characteristics associated with a conversation of the call;
a third component that is configured to analyze the set of features using multiple classifiers, wherein the multiple classifiers are generated based on multiple features extracted from a set of historical call data, wherein each of the multiple classifiers indicates one of multiple outcomes of a conversation; and
a fourth component that is configured to generate, based on the analyzing of the set of features, an on-call guidance to increase a probability of a specified outcome of the call.
27. The system of claim 26 , wherein the fourth component is further configured to:
generate a notification that indicates the participant to adopt, desist, or persist with a conversation characteristic to increase or decrease the probability of the specified outcome.
28. The system of claim 26 , wherein the fourth component is further configured to:
determine that the set of features changed as the call progressed, and
adjust the on-call guidance to indicate the participant to adopt a specified conversation characteristic that is different from the conversation characteristic.
29. The system of claim 26 , wherein the third component is configured to analyze the set of features with a subset of the multiple classifiers that include one or more features with a prediction power exceeding a specified threshold, and wherein the fourth component is configured to present values of the one or more features as the on-call guidance.