Today we’re excited to announce Alexa Conversations, a new deep learning-based approach that developers can use for creating natural voice experiences on Alexa with less effort, fewer lines of code, and less training data than before. This new model helps developers create natural, flexible dialogs within a single skill and in the upcoming releases brings multiple skills into a single conversation. Alexa Conversations is now available in developer preview in the US. Apply for the preview here.
Alexa Conversations combines an AI-driven dialog manager with an advanced dialog simulation engine that automatically generates synthetic training data. You provide API(s), annotated sample dialogs that include the prompts that you want Alexa to say to the customer, and the actions you expect the customer to take. Alexa Conversations uses this information to generate dialog flows and variations, learning the large number of paths that the dialogs could take.
Alexa Conversations helps you quickly create interactions that are more conversational, and not forced into strict patterns or sequential workflows. In the past, developers scripted every potential turn, built an interaction model, managed dialog rules, wrote back-end business logic, and analyzed logs to test and iterate. For example, the Atom Tickets skill used 5,500 lines of code and nearly 800 training examples.
Now, you provide dialog samples and Alexa Conversations predictively models the dialog path using a deep, recurrent neural network. At runtime this neural network takes the entire session’s dialog history into account and predicts the optimal next action or step in the dialog, improving accuracy and reducing your design and code efforts. It is trained to interpret dialog context in order to handle multiple user workflows, accommodate natural user input (like out-of-sequence information or corrections), address common business transaction errors, and proactively recommend additional API functionality. The Atom Tickets skill built with Alexa Conversations shrank almost 70%, to just 1,700 lines or code, and needed only 13 customer dialog samples.
Customers are asking Alexa to help with more sophisticated use cases such as ride booking, shopping, ticket purchasing or making reservations, and are expecting to converse with Alexa fluidly across multiple topics. Our vision is that Alexa learns to predict your customer’s true goal from the direction of the dialog and offers up the appropriate skills through Alexa Conversations during the natural flow. At re:MARS 2019 we showed a demonstration constructed with Atom Tickets, IMDB, OpenTable, Uber, and Amazon Pay that assembles existing APIs on the fly, helping customers plan a night out.
Developers in the preview will have access to Alexa Conversations for AI-based dialog management in existing or new skills and become eligible for early access to upcoming cross-topic capabilities. You can apply to participate in the Alexa Conversations preview by telling us about your use case in this short survey, and we’ll notify you if your application is selected. Read more about Amazon’s vision for conversational AI in the science team's blog. Read about Alexa Dialog Management here.