Human Conversation Dynamics

A shiny app that visualizes human conversation dynamics based on a 10-year subscription of 1% Twitter stream.

Go here for a publicly hosted instance of this app: https://zzeyu42.shinyapps.io/human_conversation_dynamics/

However, since I am using a free plan of the hosting service, the monthly uptime is limited (25 hours). Alternatively, you can run your local instance of the app, which should significantly boost the performance of the app. It is also easy to run it locally, especially if you're an R user (See How to run on my local machine? for instructions).

What I did

1. Extracted every retweet from the 10-year Twitter dataset;
2. Used Nominatim to geocode every retweeting/retweeted user's location, if available;
3. Used that location and the geographical information contained in the tweets, if available, to determine the countries (regions) of the retweets and retweeted tweets;
4. Counted the number of retweeting behaviors from each country (region) to another for every year;
5. Visualized the above information with this Shiny app.

How to run on my local machine?

This is a Shiny app written in R. Make sure you have an installation of R on your computer, and install the packages shiny, shinydashboard, leaflet, DT, data.table.

After that, clone this repository, and run from the command line (replace path-to-repo by the path to this repository):

R -e "shiny::runApp('path-to-repo')"

The app is now running on your local machine. You should see a browser pop up. If not, manually copy the link given by the command line interface and open that link in a browser.

Data sources

1. Twitter data: Archive Team: The Twitter Stream Grab
2. Shapes of countries (regions) (stored in data/countries.RDS to accelerate the loading process): World Bank Official Boundaries