We are drowning in a world full of unclassified text due to lack of training data. In a world, where mass text classifiers require mass acquisition of accurate data, creating custom text classifiers does not scale. We wanted to create an idea that would crowd-source the creation of this training data, allowing any users regardless of ability or knowledge of classifiers to participate.
What it does
There are two user-facing parts to the application: a Tinder-esque web app, and a chrome extension. Users of the chrome extension can highlight text on a web page and immediately submit it to a pre-defined class. These classifications are then fed to the Tinder-style app, where users are presented with pieces of text from a certain class and they have to decide whether it is correct or not.
In the future the app will provide reward points for users that select the class that is chosen by the majority of people, providing them with a motive to use the app. We're imagining someone sitting down when they're dilly-dallying waiting for a bus, and opening up Pidgeoniser to earn a couple of quid.
How we built it
Node.js and MongoDB backend; Reactjs and chrome extension user-facing frontends.
Challenges we ran into
We had to reduce the scope of the project due to time limitations.
Accomplishments that we're proud of
Produced a polished proof of concept MVP.
What we learned
We learnt how to use Docker for shipping our development environment across our (initially) Digitalocean and then AWS backends.
What's next for Pigeoniser
We are going to found a Unicorn.
Sponsored by Huel