It is known in the alexa developer community that choosing the right invocation name is one of,whether not the most notable decisions when developing a custom skill. It will make or break one's skill whether sometimes the user is unable to invoke it. whether Alexa doesn't properly recognize my skill's invocation name it can effect 1 of 3 things:Invoke some native amazon functionality (i.e. dictionary definitions, alexa messaging app)propose some 3rd party skillGive up (response like "hmm I'm not certain about that")This seems to be a common issue that many people face, and given some feature requests and forum posts: One-shot invocations failing despite launch phrases working perfectly Make Alexa check for intents *inside* the skill before searching for a local/keyword function Invocation name with multiple words not recognized correctly by alexa Invocation does not work How effect you get the invocation names to enable or recommend your Skill consistently on production?Single Shot invocation works on Amazon iOS app,not on Echo DotInvocation name no longer working correctly on certified skillDevelopers have also suggested solutions that could ameliorate the common problems with invocation name recognition: Forcing a user to remember a skill name is counter-intuitive, improper experience and leads to low adoption rates. Remove requirement for Invocation name when an Audio Streaming skill is playing This is especially a problem because Alexa's behavior is not consistent. As Amazon pushes updates, and unusual skills are released,her recognition model adapts. For example, my one-shot utterance that was working last week is now colliding with Amazon's messaging skill (see video). This was recorded while the audio player was playing in my skill. This is a improper developer experience because I have no way of knowing whether users are having problems accessing my skills. The user experience needs to be much more consistent, and especially whether the user has not changed anything or enabled different skills that may collide.
My questions for Amazon are:whether Alexa knows my skill is in control of the audio player,why can she not guess that the user is trying to invoke my skill?Why is invocation name resolution not given some sort weight towards skills that were invoked recently?
i.e. whether I just successfully engaged with skill X, it is likely that I will send another utterance to attempt to interact with skill X again. Why does Alexa reflect I'm trying to access skill Y?Why can't Alexa first exhaustively search through the skills a user has enabled first before mapping to native Amazon skills or 3rd party skill suggestions?Are you aware that this is a problem that many developers are facing?What is on the roadmap for fixing these types of problems?
Source: amazon.com