Amazon’s Alexa becomes more conversational with its new deep learning models

Amazon's Alexa is now powered with deep learning model.


On Wednesday, Amazon introduced certain extensive learning based approach for developers to get normal sound experience when they are working on Alexa. The application still in preview, allows developers to create natural skill conversation describing in short sentences of code and limited training data, Amazon describes.

Alexa, Amazon’s amicable virtual supporter is already occupied in the hearts of millions of people with its service of home interaction through cylindrical smart home speaker. “Skills,” Amazon called its set of support or the assistance that a user receives from Alexa.

In order to make more active and more advanced along with proper functioning the developers provide API access. To improve all the conversational skills of Alexa they added more items with its “skills’ functionality. A set of conversations and questions already loaded, by which Alexa would say to the user to say certain things. The user has to behave according to the commands from Alexa. Alexa works and generate dialogues as per the input data and it finishes with alexa’s conversations. Rohit Prasad, Alexa vice president and head scientist, at Amazon says that “It’s way easier to build a complex voice experience with Alexa Conversations due to its underlying deep-learning-based dialog modeling.”

“With this new approach, Alexa will predict a customer’s latent goal from the direction of the dialog and proactively enable the conversation flow across topics and skills,” Prasad pointed out. “This is a big leap for conversational AI. Our objective is to shift the cognitive burden from the customer to Alexa.” Prasad explained in a conference in Las Vegas on Wednesday.

The vice president also noticed that ML-based concept that used in Alexa to set the ‘Skills’ that allows performing multiple tasks in a single conversation. It supports the user to explain the ideas to Alexa in a simplest way and with minimal questions Alexa can work properly and adequately. It can reduce the unclearness, and save the time of the user, also it can avoid the delaying of output.

Prasad also added that, according to the current system, Alexa can manage only one-off request. It raised the need of more digital assistance to draft multiple complex questions individually. The multi skill experience is possible through the implementation of set of AI modules, combining together to create responses to the user’s queries and requests through a single dialogue. But it seemed to be a great assistance by the digital units to frame a single question that should appropriate for the user needs. That can be executed with a single individual question, which guides Alexa to a group of actions that may meet the requirements and expectation of the user. It will be the best creative output from the house of Amazone, Prasad noticed.

These days deep learning technologies is being used by top companies around the world to develop highly intelligent applications such as self-driving car, machine translations, computer vision and many others. Technologies giants are buying small machine learning startups to increase their Artificial Intelligence and Deep learning technological capabilities.

Do you want to learn machine learning and deep learning? Check TensorFlow 2.0, Artificial Intelligence, Deep Learning and Data Science tutorials.