Some Thoughts About Alexa-Cortana Friendship Part II: Group Conversation Patterns
Hello there, we are back after a brief interruption due to hurricane Irma. Today I would like to pick up where we left last week and cover the second part of the essay exploring the implications of the partnership between Amazon and Microsoft to integrate their respective digital assistants [Alexa and Cortana]. In the first part of this article, we discussed the basic integration model proposed in the first iteration of the partnership as well as a series of implications from a market perspective. Today, I would like to focus on some natural language understanding(NLU) patterns that we might see in future versions of Alexa-Cortana integrations.
As we discussed in the previous article, the initial version of the integration between Alexa and Cortana is based oin a simple patter that allows a user to launch a digital assistant[DA] (Alexa or Cortana) while interacting with the other DA. For instance, a user can ask Alexa to start up Cortana saying something like “Alexa, please open Cortana”. While this interaction pattern is certainly exciting, it results relatively simple to mimic human group conversations. As the integration between Alexa and Cortana progresses, we are likely to see more complex conversational patterns that better resemble human group interactions.
Alexa-Cortana Integration and DA Group Conversations
The integration between Alexa and Cortana can have profound implications for NLU applications as, for the first time, we have the opportunity to interact with multiple DAs as part of the same conversation(not just now but hopefully we will get there :) ). If we use human group conversations as an inspiration, there are some interesting patterns that can be followed to enrich Alexa-Cortana conversations. Let’s explore a few:
1 — Delegated Questions
The DA delegated question pattern allows a DA to ask a question to another DA on behalf of its user. Imagine a user in a dialog with an Echo device that requests some information from Cortana asking something like: “Alexa, please ask Cortana to check the number of active opportunities in our CRM”. In that scenario, the user will receive the answer from Alexa without being forced to switch DA runtimes.
2 — DA Context Sharing
Imagine take a scenario in which a user is simultaneously interacting with two DAs which share contextual information about the dialog in order to provide more accurate answers. Let’s take the following example:
User: “Cortana, what times are available in my calendar on Friday after.”
Cortana: “Currently, you have 3–5pm open”.
User: “Alexan, could you call Dr. Anderson and book an appointment for that time? “ .
In the previous dialog, Alexa and Cortana are sharing contextual information that includes the user’s calendar availability which is relevant to both parts of the dialog. I know this scenario is a bit futuristic but hopefully it illustrates the potential.
3 — DA-to-DA Conversations
If Cortana and Alexa become aware of each other’s skills and strengths, it could be possible to develop conversation in which one DA “intuitively” asks the other for information without relying on an implicit user instruction. Imagine a scenario in which a user asks Alexa for the sales forecast for the current quarter and the Amazon DA automatically asks Cortana for the marketing pipeline related to that forecast, that lives in Dynamics365, in order to provide a more complete answer.
There are many other group conversation patterns that could be relevant to multi-DA conversations. The integration between Alexa and Cortana hopefully becomes the first step towards richer group conversations in DA runtimes.