[Software] Alexa already predicts what users want based on their questions

[Dark]

Amazon has implemented a new machine learning system in its Alexa digital assistant that allows the assistant to know what users want based on their questions in order to anticipate their questions.

 The company has ensured that it seeks that users can interact with Alexa in a natural way, as if they were doing it with another human. In this sense, Amazon has added to Alexa the ability to infer the latent goals of users, without them expressing them directly and based on their questions.

   With Amazon's new digital assistant machine learning system, for example, when a user asks how long it takes to make tea, Alexa will answer, "Five minutes," and also ask, "Do you want me to set a timer of five? minutes? "

   Amazon said in a statement that detecting latent targets and formulating them into actions that often span different abilities requires a number of sophisticated algorithms.

   To determine whether to suggest a latent goal, Amazon uses a deep learning-based trigger model that takes into account the context of the dialogue, such as the text of a user's session and whether the user has engaged with Alexa's suggestions in the past.

   At the moment, this new machine learning system is only available to users who use Alexa in English in the United States.