Select Page

I described in a previous post how we can extract a network of followers on Twitter, and identify the communities to which they belong. This step allowed me to identify 6 communities:

  • Geographers,
  • a community from the Parisian Economy,
  • a group of girls who are coding,
  • a French Tech community,
  • a group of Berlin start-ups

Nothing surprising in these results: these communities describe my professional chronology.

I wanted to launch an NLP algorithm on the tweets of each of these communities and see what differentiates them. Unfortunately, I foolishly stumbled on the language barrier, so it’s hard to produce a clean corpus with tweets in German, English and French. I have simplified the problem and I used simple wordclouds. It describes as well what animates each community.

Here is my recipe on R to produce such graphs:

Extract a corpus of tweets

We extract here the last 500 tweets of our followers, published over a period of one month. This part is not super elegant because I have not found a solution to extract the tweets over a given period. The UserTimeline function from the TwitteR package only provides the ability to extract a predefined number of tweets. Maybe you have the solution?

 

Create a Twitter corpus for each community

 

Clean the corpus

 

Create wordclouds for each community

 

Results

For each group, I have isolated the subnetwork, and produced a visualization using the Gephi software. The visualizations produced are from an extraction realized between August and September 2017.

 

Frenchweb // 6032 tweets

wordcloud,twitter community, twitter followers,twitter community, twitter, audience, followers networks, folllowers

GirlsWhoCode // 1835 tweets

wordcloud,twitter community, twitter followers,twitter community, twitter, audience, followers networks, folllowers

The Berliners // 667 tweets

… Who worked in a joyous and relaxed atmosphere this summer in despite of the filthy weather #happydance 

wordcloud,twitter community, twitter followers,twitter community, twitter, audience, followers networks, folllowers