Surfacing the Nature Data You Need With a Knowledge Graph
Session Description
Climate Policy Radar opened by introducing itself as a nonprofit that began with a climate focus and is now expanding into nature, bringing together fragmented text-based data on biodiversity policy, law, finance and UN submissions into a single searchable database. Anya Schlich-Davies outlined the organisation's current database of 35,000 documents and described plans to launch a dedicated Nature Policy Radar ahead of CBD COP17 in October. Arminel Lovell then walked through how the underlying knowledge graph works, explaining how concepts are defined with synonyms and organised into parent-child hierarchies as well as looser associative connections, and ran a live demo searching CBD submissions for mentions of indigenous people, showing how the tool surfaced documents using region-specific terms such as Maori or First Nation that a plain keyword search would miss. Kyra Prins spoke to the work behind maintaining and updating the document pipeline. Following the demo, attendees split into table groups to explore the live database themselves before sketching their own knowledge graphs on paper, mapping hierarchical and associative relationships between concepts relevant to their own work.
Speakers
Anya Schlich-Davies, Nature Programme Manager, Climate Policy Radar
Arminel Lovell, Senior Knowledge Manager, Climate Policy Radar
Kyra Prins, Policy Officer, Climate Policy Radar
Watch the Session Recording
Key Takeaways
Climate Policy Radar's database currently holds 35,000 text-based documents spanning climate and nature policy, law, litigation and UN submissions, and is free to use without a login.
The organisation is expanding its knowledge graph from a climate focus into nature-specific concepts, with a first phase of Nature Policy Radar planned around CBD COP17 in October.
The knowledge graph combines hierarchical parent-child relationships between concepts with looser associative links, built to reflect the reality that concepts such as wetland restoration can plausibly sit in more than one category.
A live demo of indigenous people mentions in CBD submissions showed the tool surfacing documents, including New Zealand's, that keyword search alone would miss because relevant passages used region-specific terms rather than the phrase itself.
Attendees raised current limits of the tool, including its focus on text rather than spatial or quantitative data, and the absence of a visual, node-based way to browse the knowledge graph itself.
Several attendees expressed interest in tools that could summarise findings across many documents at once, rather than requiring manual review of individual search results.
The team is seeking input from nature tech practitioners to help shape upcoming nature-focused concepts and taxonomies, and expects to make data available via CSV export and eventually an API in the coming months.
