Partial automation of text chat conversations
To allow the human customer service agents to specialize in the instances where human service is preferred, but to scale to the volume of large call centers, systems and methods are provided in which human agents and intelligent virtual assistants (IVAs) co-handle a conversation with a customer. IVAs handle simple or moderate tasks, and human agents are used for those tasks that require or would benefit from human compassion or special handling. Instead of starting the conversation with an IVA and then escalating or passing control of the conversation to a human to complete, the IVAs and human agents work together on a conversation.
1. A method for active learning for partially automating a text chat conversation, the method comprising:
receiving a text conversation between a user computing device and a corresponding agent computer device, herein a sample;
determining an intent of the sample using a trained language model, wherein in response to the trained language model finding the sample intent indeterminate:
requesting a reviewer input to label the intent of the sample;
adding the labeled sample to a labeled training set; and
retraining the trained language model on the labeled training set; and
using the retrained language model in a text conversation between the user computing device and an Intelligent Virtual Assistant (IVA).
2. The method of claim 1 , wherein the trained language model is configured to determine an intent from the text conversation from a plurality of intents and to categorize by the trained language model the plurality of intents into intents to be responded to by the IVA and intents to be responded to by an agent, wherein a user intention is an intent which is a derived semantic representation of language of the text conversation.
3. The method of claim 2 , further comprises:
passing the text conversation to an agent of the corresponding agent computing device when the IVA recognizes the intent indicates to pass the text conversation to an agent.
4. The method of claim 3 , further comprising:
passing the text conversation from the agent of the corresponding agent computing device back to the IVA pursuant to receiving an instruction from the agent, wherein the agent labels an intent of the text conversation;
adding the labeled text conversation to the labeled training set; and
retraining the trained language model.
5. The method of claim 1 , wherein passing the text conversation from the agent of the corresponding agent computing device back to the IVA is performed seamlessly without an indication to the user computing device or to the user.
6. The method of claim 1 , further comprising sending notifications to the agent computing device, wherein a notification comprises on-screen automated assistance, work-queue entries, and reporting and post-call analytics.
7. The method of claim 1 , further comprising monitoring the text conversation between the user computing device and the IVA and adjusting the labeled training set based on the monitoring.
8. The method of claim 7 , wherein the monitoring is performed continuously during the text conversation.
9. The method of claim 1 , wherein the language model comprises a plurality of machine learning models.
10. A system for active learning for partially automating a text chat conversation, the system comprising:
an agent computing device with human agent text conversation capability and intelligent virtual assistant (IVA) text conversation ability;
a natural language understanding (NLU) component configured to use a language model in a text conversation between a user computing device and an IVA; and
an active learning engine configured to:
requesting a reviewer input to label an unlabeled sample;
add the labeled sample to a labeled training set; and
retrain the trained language model on the labeled training set.
11. The system of claim 10 , further comprising an engagement management module configured to assign an instance of the IVA to an agent or a group of agents, and to select the text conversation from a queue of live chat conversations.
12. The system of claim 10 , wherein a user intention is an intent which is a derived semantic representation of language of the text conversation.
13. The system of claim 10 , wherein the language model is configured to categorize the plurality of intents into intents to be responded to by the IVA and intents to be responded to by an agent, and to pass the text conversation to an agent of the agent computing device when the IVA recognizes the intent indicates to pass the text conversation to the agent.
14. The system of claim 13 , wherein the text conversation is passed to a human agent of the agent computing device without providing an indication to the user computing device that the text conversation is passed to the agent computing device.
15. The system of claim 10 , wherein the language model comprises a plurality of machine learning models.
16. The system of claim 10 , wherein the active learning engine is further configured to receive a plurality of unlabeled samples from a stream of unlabeled samples.
17. The system of claim 10 , wherein the NLU component is further configured to monitor the text conversation between the user computing device and IVA, and the active learning engine is further configured to adjust the labeled training set based on the monitoring.
18. The system of claim 17 , wherein the monitoring is performed continuously during the text conversation.
19. The system of claim 18 , wherein an agent computing device receives a notification in response to the monitoring, wherein a notification comprises on-screen automated assistance, work-queue entries, and reporting and post-call analytics.
20. The system of claim 10 , wherein the NLU component comprises an intent determination engine configured to determine an intent of the text conversation from a plurality of intents.