Similar to how we cannot decide for ourselves on questions such as "Where should I get food?", or "What should I do for today?". Now we can outsource our decision-making to a smart home.
What it does
Utilizing historical user input as a reference point, we have a profile on the user's likes and dislikes. When a command is issued to Google Home, tone is first analyzed with IBM Watson's NLP API, and based on the tone, a list of relevant words are used to scrape the news, with the Google News API. This is to get a current list of topics relevant to the user's preferences. Finally a decision is formulated by the multiple factors considered previously.
How we built it
We tested our code for proof of concept in Python, then translated and executed them in Node-JS, because it was universally accepted for all the platforms we used. The whole project is operated from FireBase, all inputs are sent through Watson, and all decisions are sent to the neural network for improvement on the weightings.
Challenges we ran into
Google Home had trouble connecting if it was more than 4 meters away from the WIFI router. Syntax problems when translating from Python to JS. Time complexity optimization issues for Watson API when it is not locally ran. Trouble in conversion of categorical data when trying to train the NN.
What we learned
Bring a sleeping bag.
What's next for No Bored