We have released a Data Visualization article that analyses the Belgian political leaders behaviour on Twitter. Who is the most productive ? Who is the most popular ? What are the hot topics over time ? We answer those questions through an interactive article mixing dataviz and texts. The project was published in November 2020 on the RTBF website, the first french-speaking media in Belgium.
Every political journalist or data visualization fan in Belgium saw our work on day one. Unusually for a country divided into two separate communities, the story was shared and received comments from both French and Dutch speaking people.
This data visualization project was a first for us at RTBF. We wanted to show in-house what you can do with this technology. The first reaction we had from our colleagues was: how did you do that? How long does it take to create this? It looked like magic to them.
Audience measurements showed that our readers stayed for a longer time on such pages than average. They want to go deeper and understand every part of the story. This is exactly what we want: have the audience spend more time to understand a subject based on high quality content.
Ultimately this project added value to our brand. It gave us more credibility and visibility in a competitive information market.
Data Acquisition : the tweets have been acquired through the official Twitter API.
Data Enrichment: we have used text classification techniques on the tweets. The classification algorithm was based on a word embedding model (word2vec – gensim) that has been trained on our full corpus.
What was the hardest part of this project?
The hardest part of this project was to find a point of view worth exploring for a large audience. There was no certainty about what we could find within this large amount of tweets. The combination of our 3 skills, i.e. journalism, text mining and data visualisation was key to come up with an interesting story.
Another challenge was related to the linguistic landscape of Belgium. The political leaders can write in French or Dutch, which complicates the automatic analysis of the text. To tackle this challenge, we applied a language detection algorithm on the tweets and we have implemented one classification model per language, with the same output classes.
What can others learn from this project?
Telling a story is great. Journalists have been using text, sounds and videos for ages. But what’s next? How do you go beyond your usual job? Data visualization is a new challenge for us. Space is free on the internet. Use it!
We have to see the web as a playground where everything is possible. In this case, the audience wants to see what’s behind the day to day political storytelling and create their own content by interacting with Data. They are happy to understand. Dataviz is a great way to reach that objective.
The only way to make a success out of it is to understand what every member of your team is doing. Communication is key. What is technically possible and what is not? Is it meaningful for the story to show the data like this? Did the data scientist see a trend that the journalist was not aware of?
As a journalist, don’t try to do the job of the front end developer. Learn to speak his language to leverage each other’s skill.