New AI-powered search engine advances technical capabilities to understand and analyse large quantities of data to drive climate action
The technical phase of the first UN Climate Change Global Stocktake came to a close at the UN Climate Conference in Bonn last week. The Global Stocktake, which is designed to assess the global response to the climate crisis every five years (and this year, for the first time), has seen together over 1,600 document submissions with insights and recommendations from governments and non-governmental organizations around the world to help drive climate action this decade.
Mandated by the landmark Paris Agreement in 2015, there are three core questions that the Global Stocktake seeks to answer: what is the current state of collective progress on climate action, where do we need to get to, and what are the immediate steps that Parties and non-Party stakeholders can take within the next 5 years to change course?
However, with over 170,000 pages of inputs submitted in a range of languages from across the world, understanding and synthesising the rich content within the inputs is a monumental task ahead of the conclusion of the Global Stocktake — only 6 months away, at the UN COP28 climate summit in Dubai this December.
At the UN Climate Conference in Bonn earlier this month, Climate Policy Radar launched the Global Stocktake Explorer: a search engine designed to enable users to quickly and easily understand and navigate all inputs to the first Global Stocktake ahead of COP28 this December.
The search engine - available at gst1.org - allows users to search the full text of all inputs and enables users to make sense of large quantities of complex data. Documents in all languages are auto-translated into English, which breaks down siloes and improves data accessibility (at a later stage, they will be translated into other languages as well). The tool also features bespoke machine-learning filters to help users quickly and easily identify references to critical climate action levers and impacts within the inputs.
The range of filters enable users to focus on specific elements of analysis and insight such as policy instruments, technologies, financial flows, and climate hazards. Users can use the filters to identify where inputs to the Global Stocktake mention policy levers and impact areas in the same passage, such as where inputs to the Global Stocktake mention both technologies (such as carbon capture and storage) and financial flows, where key elements of the just transition appear within specific submission types such as NDCs, or where economic policy instruments are being referenced in relation to specific climate-related hazards such as flooding and heat waves.
These filters have been created using machine learning classifiers trained using the climate-specific dataset of the Global Stocktake inputs, and by defining taxonomies taken from expert third-party sources and internal expertise. For example, within the ‘fossil fuels > oil’ sub-filter, ‘fracking’ is captured as a relevant keyword, with hydraulic fracturing, hydrofracturing and hydrofracking also included as synonymous terms to capture as much relevant content as possible.
As another example, within the ‘vulnerable groups > children’ sub-filter, child, children, youth, and young people are all included and relevant passages containing these words would be identified. The filters also use linguistic rules which capture different linguistic expressions, such as present/past tense and singular/plural, so that for example ‘flood’, ‘flooded’ and ‘flooding’ would all be returned when using the filter to explore mentions of extreme weather events within the inputs to the Global Stocktake.
As we shared the prototype search interface with attendees in Bonn last week, we received much positive and constructive feedback from national governments, NGOs, and research institutions on how the Global Stocktake Explorer can support efforts to understand and analyse critical next steps for climate action.
“This is great. We have all been wondering how everyone is going to access all this information, and this seems like it.”
At the demo session opened by the Global Stocktake Technical Dialogue Co-Facilitators Harald Winkler and Farhan Akhtar on Thursday 8 June, attendees showed considerable interest in the functionality and use cases for the search engine and raised opportunities for further enhancement. Recommendations included other languages which the documents should be translated into to further accessibility, other document types which would also be useful to include, and other filters and analytical frames for the inputs such as the Paris Agreement articles and temperature goals. We plan to address this and other feedback in the coming months.
“I love the auto-translation! This makes a very big difference.”
As the Global Stocktake moves now into the political phase ahead of COP28, we welcome more constructive feedback on how this resource can be improved to best support the process and inform efforts to enhance climate action this decade.