This project is based on the automatic analysis of news related to food security issues in French language media (national and local).
Text-mining network analysis tools are used to identify the key themes discussed in the press at a given period. Each article is then associated with these automatically reconstructed topics wether they correpond to concerns expressed at the local level, or general statements and action at national/international level.
These news items are also geo-located both by the origin of the story and the places mentionned in the story enabling to map how a given theme or issue is distributed over the world.
Moreover, themes identified at successive time steps are reconnected into streams of content. A stream visualization illustrates how topics articulate through time.
An online interface allows to visualize these maps, themes and news entries and to answer questions such as : Is an issue - concerning for example the impact of climatic change on food security - attracting more attention with time? How this specific issue relates with contiguous subjects (use of biofuel for example) ? Does the climatic change issue observed at a given time stem from, possibly various, past issue framings or is it a completely emergent topic ?
- Chavalarias, David
- Cointet, Jean-Phillipe
- Cornilleau, Lise
- Duong, Tam Kien
- Mogoutov, Andreï
- Villard, Lionel
- Roth, Camille
- Savy, Thierry
This project has been made in collaboration with UN Global Pulse.
- Complex Systems Institute of Paris (ISC-PIF)
- Ecole Polytechnique
Media Frames Streams: Monitoring World Food Security Issues
The web interface have been built with a wide range of opensource software that allowed us to build the current full functionnal website in less than one month.
We used tilemill to design the exact map we wanted to and tilestream to deploy it on an self owned map streaming server. We got the choice between polymaps and modestmaps. We choose the former one because we were planning do add multiple information layers.
The user interface rely heavily on backbone.js. The MVC cost entry was a lot worth of it and is the keystone to the quick deployement of complex interactions between the different components.
The main tube data visualization is using the paperjs library. It have been chosen for both its simplicity (compared to direct canvas js code) and its performance while keeping a vector graphic framework.
Big thanks to all opensource software developpers and the great job. We cannot imagine how many man hours all this represents and how many we should have spent if all this didn't exist.