To ensure the user phrases are properly matching to the skills, it’s important to train the Luma with periodic user data. Training is often referred to as 'Supervised training' since it involves training the NLP engine on how to better understand the user's language. The more phrases that the NLP engine has for a skill, the better it matches the phrases and even matches the skill to phrases that do not exist.
This article deals with the following:
Child pages (Children Display) page Training your Bot
This article helps you improve the accuracy in matching user phrases to skills and thereby improve the customer experience. To access Training click 
icon from the left panel. Training consists of two components Unmatched Phrases and Aborted SkillsPhrases, Aborted Skills and Suggested Training.