Amazon plans to strengthen Alexa’s capability by releasing the Alexa Skills Kit Built-in Library containing hundreds of new commands in the U.S. over the next few weeks.
In the Alexa coding language, “intents” are used to understand the basic commands from users such as “Stop” or “Cancel,” meanwhile, “slots” are used to identify information like dates and numbers. Until earlier this year, Alexa had about 15 intents and 10 slot types.
In its latest announcement, Amazon said it will roll out a library of hundreds of new intents and slots as part of the Alexa Skill Kit in the coming weeks. The new slots and intents come from Alexa’s learning of natural language over the last year.
Amazon provides these new intents and slot types as built-in so that developers don’t need to put sample assertions to train Alexa to understand user intent.
In one Amazon example, if developers want to built up a skill allowing users to ask for the temperature in the next three days, they would have to create an interaction model which included a combination of built-in and custom intents for reacting on how the question would be asked. This might involve built-in slot types for city and states, a built-in slot type for the number of days and some sample assertions to ensure Alexa will correctly understand the question each time.
With the new built-in intents library, Alexa already knows lots about weather, its attributes and in which ways a person might ask for the weather. In other words, developers are not required to create sample assertions to train the skill now.
Amazon also said that the new coming intents and slop types include books, videos and local business.
Amazon had opened up its AI capabilities as services which include the Alexa bot framework Lex. The company has also partnered with accelerator Techstars to debut a 13-week Alexa Accelerator to boost early-stage companies with a focus on Alexa towards fields such as connected home, wearables and hearables, enterprise, communication devices, connected car and health and wellness.