One of the most important parts of machine learning is how we converse with applications/devices. For example Amazon's Alexa provides a conversational interface to the home, Apple's Siri, Google's Assistant or Facebook M create a method to speak to your phone like an assistant. Each of these all allow for methods to speak and message via an "automated" experience. Facebook created their chatbot messenger framework to produced some interesting "AI" assistants which can schedule meetings or book flights. A challenge for these systems is determining a common framework of language interaction.
One method to explore is conversation flows; ask yourself how does a person converse for a short period of time and what allows for the fastest interaction.
For an example lets look at asking a persons about their sleep in the morning.
This conversational flow allows for cyclic decision making which means you can continue to provide information to the user or exit at any time.
You may ask how was this discovered... this was done by breaking down common conversations in the morning. The first step was to look at the question of "how did you sleep", followed by asking what are some of the most important pieces of information you ask yourself in the morning; e.g. weather and news.
A few interesting pieces of data which we found to provide an improved experience:
i. message question before 6am so the notification is not missed
ii. defined multiple responses to create variation day to day
iii. optimize the first response in a list to the individual user
If you would like to try out this conversational experience please do so on Facebook Messenger.
and if you have an android smartwatch try out our beta app here..
Entrepreneur in machine learning and sensor based software development (>10 years). Background in engineering, tech with professional experience in North America and Europe.