Duomoji - Track and visualise Emoji pairs usage on twitter in realtime

Poobesh Gowtham
Sep 2, 2019 · 4 min read

Check out the live version at http://duomoji.herokuapp.com
( loading might take a while as its hosted on a free server currently )

Check out the live version at http://duomoji.herokuapp.com

Whats better than an Emoji ? Two Emojis !

mojis have become a part of every conversation. Right from casual conversation on WhatsApp to official emails, emojis are everywhere. Even brands that were seen as conservative an traditional are now diving into the emoji game to stay relevant. Emojis try to capture the non textual cues of a conversation in digital text format. Nothing captures Sarcasm like 😏. Feeling annoyed ? We shoot a 😑. Want to add a lighter tone to a text ? 😛 got your back. We are at the point where even entire conversation happen with emojis.

But emojis aren’t simple creatures. There are about 800+ emojis according to the Unicode standard, but a practically infinite list of human expressions. These seemingly simple characters often pair up to express complex emotions. 😂😭 shouts something completely different from just 😂 or 😭. I set out to track and visualise that exact thing ‘How we combine emoji’s to create a new expression’ ? And what better playground than twitter ?

How we combine emoji’s to create a new expression ? And what better playground than twitter ?

As I expected, twitter was overflowing with emoji pair usage. Now I needed a system to track and analyse them.

Twitter was overflowing with emoji pair usage

Building the framework

Twitter streaming API allows tracking for upto 400 keywords. Since there are 800+ emojis, that translates to 6,40,000+ emoji pairs. So I set up a listener, that recorded 400 pair usage over a period of 1hour at a time, and aggregated the total of all pair usage over a period of 3 weeks. I decided to track the top 400 of such pairs. A fairly straight forward system was formulated and deploye on a cloud server.

Basic Architecture of Duomoji

Now with the system tracking the top emoji-pair usage, I set out to build the UI. I wanted to visualise how each of the emojis pair with each other. I found that only a few combinations were used very frequently. So that reduced the need to visualise 400 pairs to just 35 pairs. I tried out a few iterations to finalise on the layout.

Iterations for emoji pair usage visualisation

I decided to go with a chord diagram as it captures the relationships in a much bettter way. I added lines that go from the first emoji of the pair to the second. The lines gave a good idea of which were the most paired emoji.

Visualising emoji pair connections

I added a live tracker that ranks the top 16 emoji pair and blinks when the pair is used by some twitter user at the moment.

Although the previous visualisation gave a good picture of what the top paired emojis are, I wanted to further see what emoji was most paired with a particular one. I used D3 to botch up a quick bubble chart with animations.

What’s paired most with what ?

There were quite a few interesting observations of how we pair up emojis. One could spend a lot of time interpreting the relationships. I’ll leave it up you to figure out the insights. Do let me know if something strikes you as very interesting!

Future Roadmap

  • Try to derive the reason behind each popular pair.
  • Create a new emoji that conveys the combined emotions of two emojis

The entire project is Open source, hosted on Github. Please fo hit me up for feedback or suggestions. Collaborations are always welcome.


Check out the live version at http://duomoji.herokuapp.com

A sky full for stars

Poobesh Gowtham’s portfolio