Amazon’s Alexa will now guess what skills you want to use based on your questions

mediumThis post was originally published by Richa Sharma at Medium [AI]

Amazon’s Alexa has always made us wonder with its latest and upgraded features to assist us in our daily tasks. The company says that it’s Artificial Intelligence assistant has more than 10,000 skills. However, not all of them are useful and used by users across the world. Those are the skills that are much harder to discover as well. Earlier in November, Amazon had announced that it has launched its new way to surface skills. it would be by guessing what users are after when they talk to Alexa about other tasks or seek help for various activities throughout the day. Well, that is a step ahead in terms of technological intelligence and performance!

Image for post
Amazon Alexa

Amazon calls this process “[inferring] customers’ latent goals.” It means that Alexa is working out on any questions that are implied by other queries. The company explains this by stating the example of a customer asking, “How long does it take to steep tea?” to which Alexa will respond with the approximate time required to steep tea, “five minutes” and further ask the follow-up question: ”Would you like me to set a timer for five minutes?” It aims at ensuring enhanced customer service with the aid of the various, brilliant features of Alexa.

This may sound very intelligent to you, rightfully justifying the very essential theme of Alexa- Artificial Intelligence. However, this feature is already available to Alexa customers in English in the United States that has increasingly become annoying for them. Imagine your smart Alexa asking you irrelevant questions based on its false interpretations. It would end up making you feel annoyed in no time.

Image for post

The company gave other examples to explain the new feature and added that other skills that are present in the system include Bark Like a Dog, Earth Cam, Harry Potter Trivia, Historical Voices, Slow Speller, Xbox, and Zoo Keeper. Although some of the enthusiastic users will make the most of it, it may become potentially annoying for other people.

The company explained in a blog post that it makes use of technologically advanced techniques like the deep learning-based trigger model that takes into account factors including the text or speech of the customer’s current session with Alexa. It also notes whether the customer has engaged with Alexa’s multi-skill suggestions in the past. Now that is how the company’s one of the smartest devices work with leaving some space of annoyance for some of you. Amazon said that “Over time, the discovery model improves its prediction through active learning, which identifies sample interactions that would be particularly informative during future fine-tuning”. Once the system starts suggesting some latent goals, Alexa uses a semantic-role labeling model to look for names and other related arguments from the current conversation that leads up to preparing and asking the follow-up question. The device also uses bandit learning that supports machine learning to track whether recommendations are helpful to the customers or not.

Recommended Read: Quick Tips To Design A Brilliant Voice User Interface

For a virtual assistant to achieve this level of intelligence is quite difficult to decipher right now. It also requires a set of sophisticated algorithms to come up with an idea so brilliantly coordinated and executed. To come up with the questions bases on the past or latest conversation, Alexa has to analyze multiple features in users’ requests and compare them to previous patterns of interaction. The device tries to learn from customers’ behaviors. Say, for example, storing in its memory that users who ask how long tea should brew for often subsequently request a timer to be set up for that amount of time as explained in the example by the company. It is even more challenging to develop a method for the device to predict the user’s next question or activity. The algorithm is designed in such a manner that tries to gather a contextual understanding of the words spoken by the user enabling Alexa to understand the intent behind the question.

Hands down to Amazon’s engineers and developers who have created an altogether intelligent device so far.

Spread the word

This post was originally published by Richa Sharma at Medium [AI]

Related posts