Dynamic topic guidance in the context of multi-round conversation
A topic guidance method, system, and computer program product for creating a conversation model by learning a conversation pattern from a customer response to a conversation topic segment and suggesting a conversation topic for the agent to engage the customer in the conversation topic based on a customer response associated with the conversation model.
1. A computer-implemented topic guidance method for a call between an agent and a customer, the method comprising:
creating, via a processor on the computer, a conversation model by learning a conversation pattern from a customer response to a conversation topic segment, a current state of the customer, and a context of the conversation topic segment; and
suggesting, via the processor on the computer, a conversation topic for the agent to engage the customer in the conversation topic based on a customer response associated with the conversation model, analyzed conversation text the current state of the customer, and the context of the conversation topic segment.
2. The computer-implemented method of claim 1 , wherein:
in a training phase of the conversation model:
historical call conversation data is segmented into conversation segment topics;
the customer response in the historical call conversation data is detected in response to each of the conversation segment topics;
construct a customer state; and
the conversation pattern is learned by learning a pattern of customer responses to the conversation segment topics in order to create the conversation model; and
in a run-time phase of the conversation model:
the suggesting suggests the conversation topic for which the agent to engage the customer; and
a real-time response to the suggested topic is analyzed to dynamically adapt the suggested conversation topic based on the conversation model being referenced for the real-time customer response to the conversation topic.
3. The computer-implemented method of claim 1 , wherein the suggested conversation topic is displayed in a decision-type tree of which multiple conversation topics are displayed with an expected customer response derived from the conversation model.
4. The computer-implemented method of claim 1 , wherein the conversation model is learned by a Markov Decision Process based on at least one of a finite set of states, a finite set of actions, and a discount factor.
5. The computer-implemented method of claim 4 , wherein the finite set of states includes an emotion, a talking time, a historical transaction, a contacted frequency, a demographic information, and a customer category, and
wherein the finite set of actions includes conversation topics.
6. The computer-implemented method of claim 1 , wherein one of a Q-learning technique and a deep Q-network is used in the creating to create the conversation model.
7. The computer-implemented method of claim 1 , embodied in a cloud-computing environment.
8. A computer program product for topic guidance for a call between an agent and a customer, the computer program product comprising a computer readable storage device having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform:
creating a conversation model by learning a conversation pattern from a customer response to a conversation topic segment, a current state of the customer, and a context of the conversation topic segment; and
suggesting a conversation topic for the agent to engage the customer in the conversation topic based on a customer response associated with the conversation model, analyzed conversation text the current state of the customer, and the context of the conversation topic segment.
9. The computer program product of claim 8 , wherein:
in a training phase of the conversation model:
historical call conversation data is segmented into conversation segment topics;
the customer response in the historical call conversation data is detected in response to each of the conversation segment topics;
construct a customer state; and
the conversation pattern is learned by learning a pattern of customer responses to the conversation segment topics in order to create the conversation model; and
in a run-time phase of the conversation model:
the suggesting suggests the conversation topic for which the agent to engage the customer; and
a real-time response to the suggested topic is analyzed to dynamically adapt the suggested conversation topic based on the conversation model being referenced for the real-time customer response to the conversation topic.
10. The computer program product of claim 8 , wherein the suggested conversation topic is displayed in a decision-type tree of which multiple conversation topics are displayed with an expected customer response derived from the conversation model.
11. The computer program product of claim 8 , wherein the conversation model is learned by a Markov Decision Process based on at least one of a finite set of states, a finite set of actions, and a discount factor.
12. The computer program product of claim 11 , wherein the finite set of states includes an emotion, a talking time, a historical transaction, a contacted frequency, a demographic information, and a customer category, and
wherein the finite set of actions includes conversation topics.
13. The computer program product of claim 8 , wherein one of a Q-learning technique and a deep Q-network is used in the creating to create the conversation model.
14. A topic guidance system for a call between an agent and a customer, said system comprising:
a processor; and
a memory, the memory storing instructions to cause the processor to perform:
creating a conversation model by learning a conversation pattern from a customer response to a conversation topic segment, a current state of the customer, and a context of the conversation topic segment; and
suggesting a conversation topic for the agent to engage the customer in the conversation topic based on a customer response associated with the conversation model, analyzed conversation text, the current state of the customer, and the context of the conversation topic segment.
15. The system of claim 14 , wherein:
in a training phase of the conversation model:
historical call conversation data is segmented into conversation segment topics;
the customer response in the historical call conversation data is detected in response to each of the conversation segment topics;
construct a customer state; and
the conversation pattern is learned by learning a pattern of customer responses to the conversation segment topics in order to create the conversation model; and
in a run-time phase of the conversation model:
the suggesting suggests the conversation topic for which the agent to engage the customer; and
a real-time response to the suggested topic is analyzed to dynamically adapt the suggested conversation topic based on the conversation model being referenced for the real-time customer response to the conversation topic.
16. The system of claim 14 , wherein the suggested conversation topic is displayed in a decision-type tree of which multiple conversation topics are displayed with an expected customer response derived from the conversation model.
17. The system of claim 14 , wherein the conversation model is learned by a Markov Decision Process based on at least one of a finite set of states, a finite set of actions, and a discount factor.
18. The system of claim 17 , wherein the finite set of states includes an emotion, a talking time, a historical transaction, a contacted frequency, a demographic information, and a customer category, and
wherein the finite set of actions includes conversation topics.
19. The system of claim 14 , wherein one of a Q-learning technique and a deep Q-network is used in the creating to create the conversation model.
20. The system of claim 14 , embodied in a cloud-computing environment.