Few business managers have the time to get into their database. Even fewer have the time to learn SQL and get important analytics without a data analyst. With VQL [Voice Query Language], common critical database info can be available by just asking google.
What it does
Ask google to give you common business database information without ever touching SQL.
How we built it
Created a flask API in a virtual environment. The flask API interfaces between the business 8 base database and the voice assistant API. The flask API includes a Natural Language Processing model along with several functions to handle query communication with the google cloud APIs and the database. NLP was built using sklearn, spacey, and pandas. Deployed the flask API on Heroku and managed version control through GIT.
Challenges we ran into
We originally attempted to run the system through Amazon Alexa by building an Alexa Skills app to communicate to the flask app. Alexa Skills wouldn't give the original user input we need to run through the flask API. We also attempted to host the flask API locally but communication was difficult through FIU's routers to the google action.
Accomplishments that we're proud of
Google assistant voice requests can deliver actual database information.
What we learned
Lots of database HTTP POST requests programming. First implementation of 8 base for everyone. How to query JSON files between APIs.
What's next for VQL
Building a frontend with text boxes of possible questions followed by text/graphical based output. Further data modeling for deeper analytics. Simple docking system for customer's database. Building Alexa skills communication. Mobile optimization.