Global Stocktake Explorer In partnership with UNFCCC logo UNFCCC

Frequently asked questions

You can use Climate Policy Radar’s Global Stocktake Explorer to:

What is the Global Stocktake Explorer?

The Global Stocktake Explorer is a research tool developed by Climate Policy Radar to support the analysis of inputs to the first Global Stocktake. These include thousands of inputs that were submitted via the Global Stocktake Information Portal, and other input documents available on other UNFCCC portals such as the National Adaptation Plan Central Portal.

The Global Stocktake Explorer allows searching the full text of all the inputs to the Global Stocktake at once. Documents in all languages are auto-translated to English (see more below). Key climate action topics under the themes of mitigation, adaptation, means of implementation and cross-cutting have also been labelled in the text using machine learning methods supervised by a team of climate policy experts and data scientists. These are reflected in the filters that you can use to understand and analyse the inputs in addition to the search functionality.

Read our methodology to find out more about how we developed the thematic filters and text labels.

Using the Global Stocktake Explorer

You can access, explore and search the full text of inputs to the Global Stocktake, in two ways:

Free text search: click on the search bar at the top of the screen and enter a search term. In the search results on the right hand side, you’ll see a series of paragraphs or documents sections containing your search term, highlighted in yellow. Each passage of text comes from an input to the Global Stocktake. You’ll see the document title, author and document type labelled above the passage.

Thematic filters: select one or more topics and sub-topics from the list on the left-hand side of the screen. Each passage of text in your search results references the topic(s) you’ve chosen, and comes from an input to the Global Stocktake. You’ll see the document title, author and source labelled above the passage. Words related to the topic(s) you chose will be highlighted in grey and labelled. For more details about how the filters work, see FAQ “How do the thematic filters work?”

You can use the free text search and thematic filters simultaneously. You can also download your search results using the ‘download CSV button’ on the top right corner of the search results pane.

What documents are searchable via the Global Stocktake Explorer?

The Global Stocktake Explorer has over 1,800 inputs to the Global Stocktake defined according to decision 19/CMA.1 para 37. These inputs have been sourced from the following portals:

The Global Stocktake Explorer contains the following input documents:

  • Nationally Determined Contributions (NDCs)
  • IPCC Assessment Reports
  • Fast-Start Finance Reports
  • National Communications
  • Biennial Reports
  • Biennial Update Reports
  • Annual Compilation and Accounting Reports
  • Facilitative Sharing of Views Reports
  • Global Stocktake Synthesis Reports
  • Intersessional Documents
  • Long-Term Low Greenhouse Gas Development Strategies
  • National Inventory Reports
  • Pre-Session Documents
  • Progress Reports
  • Publications submitted via the Global Stocktake Information Portal
  • Reports submitted via the Global Stocktake Information Portal
  • Statements submitted via the Global Stocktake Information Portal
  • Submissions to the Global Stocktake
  • Summary Reports
  • Synthesis Reports
  • Technical Analysis Summary Reports
  • Adaptation Communications
  • National Adaptation Plans
  • Technology Needs Assessments

We plan to update the Global Stocktake Explorer to include additional inputs and relevant documents during summer 2023.

Is the data up to date?

The documents were last downloaded from the UNFCCC portals on the 24th of May 2023, after the March 2023 cut-off date for inputs to the Global Stocktake.

How do the thematic filters work?

The majority of filters have been created using bespoke taxonomies of each topic, containing categories and keywords to capture relevant content for each topic. We indicate where a certain keyword may have multiple synonyms or different linguistic expressions, such as present/past tense, singular/plural, and acronyms, and all of these variations are automatically included.

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.

There are three exceptions to this: the ‘financial flows’ filter uses a combination of keywords and a machine learning model to detect mentions of money (e.g. “$1 million”), and the ‘sectors’ and ‘policy instruments’ filters use machine learning classifiers which operate on each text passage. You can find more details in the methodologies for the corresponding taxonomies.

The taxonomies were created either by using existing taxonomies developed by expert third-party sources or compiled based on our own internal expertise and desk research. Read our methodology for more detailed information.

How were the filters decided and can you add more?

The filters were determined through extensive consultation with the UNFCCC and other expert groups, in combination with desk-based research.

We intended to improve our current filters and add new ones in due course. If you have any suggestions for improving existing filters or new ones that you would like to see included, please get in touch through our Contact page.

How are the search results prioritised?

Search results are prioritised using the default prioritisation method of Opensearch which uses the following approaches:

  • How many of the search terms appear in each result - the more terms that appear, the higher the rank.
  • How many times each of the terms in the search query appears in each result - results that mention the terms in the search query more frequently will be prioritised over those that don’t.
  • How frequent the term is in the whole text dataset - rarer terms from a search query are assumed to carry more “signal” and are therefore more important for ranking. For example, a term that appears in 9% of search results provides less information for ranking than one that appears in 5% of results.

How are the documents translated to English?

Documents are translated to English using Google’s Cloud Translation API. While the quality of auto-translation does not always capture full meaning and nuance from the original language, we hope it serves as a first step to improve accessibility. We are planning to add auto-translation to other languages in the future, as well as to improve the quality of translation using other, potentially domain specific, models.

Does the Global Stocktake Explorer use AI?

Some topics (concepts) in Global Stocktake Explorer use classifiers, a machine learning technique that trains a computer model to learn what kind of language relates to a particular topic (concept), such as sectors or policy instruments. We trained our algorithms using the dataset of the Global Stocktake inputs that were labelled and reviewed by our policy team.

Using these approaches, it is not possible to ensure that we have found every instance of relevant language (known as recall), or ensure that every piece of text we identify is definitely related to a concept (known as precision). We are continuously working to improve the recall and precision of our approach but please be aware that not all relevant text may be identified and that some results identified may be false positives.

Climate Policy Radar developed the Global Stocktake Explorer by combining our team's policy and technical expertise with machine learning methods. This "augmented intelligence" approach ensures accuracy of the algorithms through careful training, review and supervision by our team.

An AI computer vision model is also used to extract the text from formatted PDFs.

Do you use large language models (LLMs)?

Global Stocktake Explorer does not use LLM technology (like ChatGPT). We’re carrying out research and development work to leverage the power of LLMs in our other products, such as to help users understand why search results have been returned.

Fundamental to our work with LLMs is ensuring we mitigate risks such as bias or “hallucinations”, in which LLMs fabricate untrue information. Any application of LLMs in our tools will be carefully designed and audited to minimise the impacts of these limitations.

What does ‘research and development preview’ mean?

A research and development (R&D) preview is a cut-down version of a tool for testing out ideas. Global Stocktake Explorer provides an early view of our R&D work to make large libraries of documents easier to search and analyse by identifying language related to taxonomies of important climate topics. Releasing it as a preview allows us to gather early feedback from our community on how well it’s working and how people use it. We use these findings to make iterations and improvements before incorporating it into our main tools.

How do I download and reference the data?

You can download your search results using the ‘Download results as CSV’ button on the right side of the page.

If you have used the Global Stocktake Explorer, please cite this source as follows: ‘Data sourced from Climate Policy Radar’s Global Stocktake Explorer, made available at under the Creative Commons CC-BY licence.’

How do I report a bug or error, provide feedback, or make suggestions?

Please get in touch with Climate Policy Radar via our Contact page. We appreciate you taking the time to do this, and will aim to get back to you within 3 days.