Persisting an AI-supported conversation across multiple channels
A method and computing platform to imitate human conversational response as a context transitions across multiple channels (e.g., chat, messaging, email, voice, third party communication, etc.) where inputs to the system are categorized into identified speech acts and physical acts, and a conversational bot is associated to the channels. In this approach, a data model associated with a multi-turn conversation is provided. The data model comprises an observation history, wherein an observation in the observation history includes an identification of a channel in which the observation originates. As turns are added to the multi-turn conversation, a conversational context across multiple channels is persisted using the data model. Using this approach, an AI-supported conversation started in one channel can move to another conversation channel while maintaining the context of the conversation intact and coherent.
1. A method for imitating a human conversational response across multiple channels, comprising:
associating a multi-turn conversational bot with one or more of the multiple channels;
providing a data model uniquely associated with a multi-turn conversation via the multi-turn conversational bot, the multi-turn conversation being between an actor and the multi-turn conversational bot, the data model comprising a linear sequence of observations, wherein an observation is an atomic speech or physical action taken by the actor, the data model also including a set of one or more events that have been determined to represent a meaning of the conversation history since the multi-turn conversation begins, wherein each event is composed of a sequence of observations and is determined based on a closest exact or inexact match to event annotations overlaid onto conversational fragments in a set of historic conversational transcripts, wherein an inexact match is associated with an aliased event having an event expression that is an approximate match to an existing event;
augmenting each observation in the sequence of observations to include an identification of a channel in which the observation originates;
as one or more turns are added to the multi-turn conversation, persisting a conversational context across multiple channels using the augmented observations in the data model; and
adapting a particular response or action provided to an end user based on the channel currently in use for a particular turn.
2. The method as described in claim 1 wherein persisting the conversational context includes transitioning the multi-turn conversation from at least a first channel to a second channel.
3. The method as described in claim 2 wherein the channels are one of: chat, messaging, email, voice, and a communication initiated by a third party.
4. The method as described in claim 2 further including setting a value of an attribute during a first turn that occurs in the first channel, and using the value during a second turn that occurs in the second channel without prompting again for re-entry of the value.
5. The method as described in claim 1 wherein respective turns in the multi-turn conversation are associated with distinct end user sessions.
6. The method as described in claim 2 wherein the multi-turn conversation transitions from the first channel to the second channel in response to an action initiated by the end user or as a result of activity associated with a process.
7. The method as described in claim 1 wherein the multi-turn conversation is associated with a particular end user.
8. The method as described in claim 1 wherein providing the data model further includes:
receiving an input that includes an attribute; and
responsive to receipt of the input, retrieving the data model from a set of saved data models when a value of the attribute is recognized as being present in the data model.
9. The method as described in claim 1 wherein the multi-turn conversation is associated with a conversation initiated between a first and second human end user.
10. A software-as-a-service computing platform, comprising:
computing hardware;
computer software executing on the computer hardware, the computer software comprising computer program instructions executed on the computing hardware and configured to imitate a human conversational response across multiple channels, the computer program instructions configured to:
associate a multi-turn conversational bot with one or more of the multiple channels;
provide a data model uniquely associated with a multi-turn conversation, the multi-turn conversation being between an actor and the multi-turn conversational bot, the data model comprising a linear sequence of observations, wherein an observation is an atomic speech or physical action taken by the actor, the data model also including a set of one or more events that have been determined to represent a meaning of the conversation history since the multi-turn conversation begins, wherein each event is composed of a sequence of observations and is determined based on a closest exact or inexact match to event annotations overlaid onto conversational fragments in a set of historic conversational transcripts, wherein an inexact match is associated with an aliased event having an event expression that is an approximate match to an existing event;
augment each observation in the sequence of observations to include an identification of a channel in which the observation originates;
as one or more turns are added to the multi-turn conversation, persist a conversational context across multiple channels using the augmented observations in the data model; and
adapt a particular response or action provided to an end user based on the channel currently in use for a particular turn.
11. The computing platform as described in claim 10 wherein computer program instructions that persist the conversational context further include computer program instructions configured to transition the multi-turn conversation from at least a first channel to a second channel.
12. The computing platform as described in claim 11 wherein the set of input and output channels are one of: chat, messaging, email, voice, and a communication initiated by a third party.
13. The computing platform as described in claim 11 wherein the computer program instructions further include computer program instructions configured to:
set a value of an attribute during a first turn that occurs in the first channel; and
use the value during a second turn that occurs in the second channel without prompting again for re-entry of the value.
14. The computing platform as described in claim 10 wherein respective turns in the multi-turn conversation are associated with distinct end user sessions.
15. The computing platform as described in claim 11 wherein the multi-turn conversation transitions from the first channel to the second channel in response to an action initiated by the end user or as a result of activity associated with a process.
16. The computing platform as described in claim 10 wherein the multi-turn conversation is associated with a particular end user.
17. The computing platform as described in claim 10 wherein the computer program instructions configured to provide the data model further include computer program instructions configured to:
receive an input that includes an attribute; and
responsive to receipt of the input, retrieve the data model from a set of saved data models when a value of the attribute is recognized as being present in the data model.