Every second, more than one hectare of tropical forestland is destroyed around the planet, in many cases, by large global corporations in the incessant hunt for agricultural space, timber, and urban expansion. Many times, this shows up on the web and in social media, but finding credible, up-to-date information from the thousands of posts produced every day can be a full-time job for a researcher, and a tough proposition for any non-profit working in this space.
MANA-Vox, an NGO that is based in Nice, France, provides data-driven, up-to-date rankings of corporations’ involvement in environmental issues, beginning with deforestation but with the potential to expand into other areas such as water and air pollution. This information can be used to shape policies: at both government and corporate levels. The goal of the organization is to provide dashboards to non-profits working in the environmental space that can benefit from the information, as well as companies and financial institutions that are lacking this breadth of information.
At its core, the MANA-Vox application uses artificial intelligence (AI) to scour information from trusted sources and categorize that information based on the model on which the system is trained.
After initially building a proof-of-concept application that is based on IBM Watson artificial intelligence components, the MANA team turned to the IBM Garage to make the application scalable. These are the core components of the system designed by the team:
- A list of trusted sources (social media accounts and web pages) that are scoured daily for any updates, which can include web pages embedded as links in social copy
- A natural language analysis engine that is trained on the domain that analyzes the content, pulling out company names and categorizing the content, running on a serverless architecture that can scale down to zero when the application doesn’t need to run
- An API-driven front end that can be connected to any dashboard or management interface
To bring the solution to life, we often worked collaboratively, sometimes using pair programming in areas like development of algorithms and building front-end views.
Scaling the app to other domains
The system analyzes over 1,000 tweets a day, looking at the social post content as well as the content in any links. Typically, only approximately 20 of these are considered relevant by the trained AI system. Doing this work manually can take hours and doesn’t scale well. With an application like MANA-Vox that uses AI, the core code does not need to be rewritten, only the training models updated to understand a new domain such as air or ocean pollution.
On the subject of scaling, we have adopted a serverless architecture using IBM Cloud Functions for identifying sources and analyzing content so that the application only runs when needed and scales to zero the rest of the time. The application can grow based on demand, and we only have to focus on app development, not infrastructure considerations.
Artificial intelligence opens up new potential for creating apps with fresh potential for enacting real social change. As data gets democratized, government and corporate policies can be better informed to help us protect the planet.
Read the case study to learn more about the ManaVox solution and their collaboration with the IBM Garage team.