Inspiration
One of our teammates receives quite a few marketing calls. Usually he ignores them. But then the automated bots simply fill up his voicemail, so it becomes a chore to simply find the messages that matter. We wanted to create a modern voicemail client that helps all clients, individuals or businesses, sift through the spam and find the stuff that matters.
What it does
Visual Voicemail 2.0 uniquely categorizes voicemails based on a machine learning algorithm that learns what messages you mark as spam and which ones you want to keep. The neural network classifies all voicemails into 4 categories (spam, not spam, urgent, and low priority) and prioritizes urgent ones while hiding spam.
How we built it
Used sci-kit learn on a python server for machine learning. UI is an android app to mimic the extant Visual Voicemail feature.
Challenges we ran into
Getting neural network to accurately categorize messages.
Accomplishments that we're proud of
Categorizes voicemails pretty well. More training required of course.
What we learned
We explored machine learning for the first time, in addition to learning more about RESTful APIs and Android development.
What's next for Visual Voicemail 2.0
The next steps for Visual Voicemail 2.0 would be optimizing the UI to allow for more user control of the filtering of their voicemails, as well as an expansion of the training sets for the machine learning algorithm to better identify and classify incoming voicemails.
Log in or sign up for Devpost to join the conversation.