CC7 - Extract guidance and actions from Climate Change documents using Machine Learning

Project start and end dates:
2019-09-01 to 2019-12-15

Goals & Objectives

We are in the design and planning phase of a new climate change impact planning software product (related to “UN Sustainable Development Goal 13: Take urgent action to combat climate change and its impacts”). The product will recommend a set of actions that ordinary citizens can take to avoid, mitigate, adapt, or rebuild from climate change disasters. Although there is a wealth of guidance information out there in PDFs or on websites, rather than go through those one by one, we want to see if machine learning NLP and NLU techniques (and others?) could be used to automatically extract that kind of guidance from a document.

We would like you to explore ML techniques on a set of PDF and website content and produce some “climate change actions” via a model. This is exploratory and so we anticipate a few iterations where you produce some outputs, we examine those, and see how to structure the results into our proposed action structure. At the start of the project we will provide a set of PDF content, some suggested sentence structures/keyword examples, and a proposed action format (in JSON) for the model, and anything else you need.

Project Outcome

The University of Toronto students investigated whether (and how) machine learning techniques could be used to extract meaningful climate change actions from existing content. While the project work was intended to be an exploration/feasibility study, the results are very polished and extremely promising.

The exceptional level of their project deliverable testifies to the quality, thoroughness, and professionalism of their research, teamwork, and problem solving approach. In arriving at their chosen “minimum viable product”, they explored many potential NLP, NLU, and other techniques and libraries.

Importantly for such a complicated domain, they were able to articulate the complex math and ML techniques and approaches in a relatively non-technical way. This allowed us to gain a (limited) understanding of what they were doing, and why. They frequently asked for feedback to adjust and improve their algorithm and align it with project goals and objectives. They also indicated many areas for future improvement or investigation.

Overall the team’s result was exceptional and of immense insight. They not only met but greatly exceed our expectations. We highly recommend them and their superlative work.


We would like to thank Rahim Jiwa, Linda Peto, and the other students and teaching staff of SCS-3253 – Machine Learning course. *

* For privacy reasons, we only list people who gave us permission to do so. Did you contribute to this project? Contact us to be added!

Related Project

Climate Change Impact Planner
When it comes to climate change adaptation and mitigation, we believe there is a critical communication gap between government officials, scientists, and experts (“trusted authorities”), and ordinary citizens. As citizens, we do not know what specific actions we can take to avoid, mitigate, adapt to, or recover from the effects of climate change in our personal situation (such as in our home and neighborhood). Even if we do act, trusted authorities do not know if we are following their guidance and cannot study the effectiveness of our actions. The anticipated product will use satellite Earth Observations, big data, and machine learning to identify past climate change impacts and predict future impacts. It will focus on end users and stakeholders (such as municipal governments or insurance companies) who need to understand and plan around climate change impacts without requiring them to know the underlying science or technology.

Academic Institution

University of Toronto

Toronto, Ontario

The University of Toronto is a public research university in Toronto, Ontario, Canada, located on the grounds that surround Queen’s Park. It was founded by royal charter in 1827 as King’s College, the first institution of higher learning in the colony of Upper Canada. 

A Riipen Project

Riipen is your online platform for virtual project-based learning

Get hands-on support from our students through an in-class project or virtual internship.

  • In-class projects allow you to connect with one of our educators to embed your project into the students’ curriculum. Become the real-life case study for students in the classroom!
  • Virtual internships are similar to in-person internships, except they are project-based with a clear outcome and the engagement is primarily done online.