Amazon reported that its 2018 holiday season delivered record sales for smart-home devices. Its Echo smart speaker featuring the Alexa virtual assistant and supported by machine learning was among the most popular items.
In a post the company released a list of its top Alexa skills and apps controlled by voice and used in the U.S. for 2018. The top 10 lists mostly popular and innovative skills spanning games, wellness, daily habits, and family oriented.
Alexa learned a huge number of new things in 2018, making it easier to search and find accurate information to build skills.
Alexa skills help people search for information, learn something new, or carry out a task such as making a purchase, which can assist brands in engaging with consumers.
Developers built more than 70,000 skills for Alexa and increased the number of smart-home devices to more than 28,000 from about 4,500 brands. The increase, in part, resulted from ways that Amazon improved on developer tools and natural dialog features, as well as added ways that developers could request permission from users to provide responses based on their location.
The list begins by naming the best game skills that include Question of the Day, Yes Sire, The Magic Door, Would You Rather For Family, Skyrim Very Special Edition, Trivia Hero, Heads Up!, Beat the Intro, Nat Geo Quiz, and World Mathematics League.
The best family skills range from how to build a lemonade stand to making brushing your teeth fun and one from Sesame Street.
Since fall 2018, Amazon began working on Alexa’s self-learning techniques that teach the VA to recover from its own errors. Early in December, Alexa AI Director of Applied Science Ruhi Sarikaya provided details of the machine-learning advances.
He wrote about refining context in how human-computer interaction through natural-language improves accuracy, and how machine-learning technology can help to perform name-free skill interaction, which allows customers to more naturally discover, enable, and launch Alexa skills.
The example Sarikaya gave shows how people can order a car from a ride-sharing service without specifically naming the company. “This requires a system that can process many contextual signals to automatically select the best skill to handle a particular request,” he wrote.