You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

Climate Policy Radar Open Data

This repo contains the full text data of all of the documents from the Climate Policy Radar database (CPR), which is also available at Climate Change Laws of the World (CCLW).

Please note that this replaces the Global Stocktake open dataset: that data, including all NDCs and IPCC reports is now a subset of this dataset.

What’s in this dataset

This dataset contains two corpus types (groups of the same types or sources of documents) which are accessible via the document_metadata.corpus_type_namefield. See the full methodologies (CPR, CCLW) for details.

  • Laws and policies: national climate laws and policies
  • International agreements: decisions by, as well as submissions to, international processes such as Nationally Determined Contributions (NDCs) submitted to the United Nations Framework Convention on Climate Change. This includes both Party and non-Party authors.

As the CPR dataset expands, this list may include other document types.

Dataset version and maintenance

This dataset contains all of the documents of the aforementioned corpus types that are available on the Climate Policy Radar app. Laws and policies are filtered to those published more than 6 months ago, and international agreements are live at the date of last publishing.

We are also working on an API for our live dataset, which will be available soon.

Using this dataset

We’ve created some demonstrations of how to efficiently load, perform analytical queries and produce visualisations on this dataset. These are available on Github at ClimatePolicyRadar/open-data.

Data model

Each row in this dataset is a text block (i.e. passage or chunk) from a document.

Individual documents are grouped into 'document families'. The way to think of families is as follows:

  • Each row in the dataset is a physical document. A physical document is a single document, in any format (mostly PDF and HTML).
  • All physical documents belong to document families. A document family is one or more physical documents, centered around a main document, which jointly contain all relevant information about the main document. For example, where a document has a translation, amendments or annexes, those files are stored together as a family.

Some families are further grouped into collections, particularly for longstanding policy responses that have developed over longer periods. An example of this is the European Union’s Common Agricultural Policy.

Column names and descriptions

Columns are split into a few different namespaces for ease of navigating, denoted by different prefixes: document_ and document_metadata for document related columns; text_block_ for text block (passage) related columns; pdf_data and _html_data for columns specific to each file format, and pipeline_metadata for some metadata that our pipeline stores about each document.

Document-related

Column Definition Sample Data
document_id and document_metadata.import_id The internal system ID used by our database developers to uniquely identify a document
document_metadata.slug The unique identifier for this document used in the document URL. adaptation-strategy-to-climate-change-in-the-czech-republic_213b
You can rebuild the document URL from this ID using the format <Domain>/documents/<Document ID>, eg. https://app.climatepolicyradar.org/documents/adaptation-strategy-to-climate-change-in-the-czech-republic_213b
document_metadata.document_title The title as displayed when a document file is opened.
document_metadata.description A short description of the document.
document_metadata.publication_ts The main date associated with this document. For the current data types this is publication date.
document_metadata.source_url The address to a web page where you can directly read or access the content of a policy document
document_metadata.type The form by which a policy document presents itself (e.g. a plan, strategy, law). Each document belongs to one of a fixed set of types defined by CPR.
document_metadata.source The data source from which the policy document was extracted. “UNFCCC”, “CPR”, “CCLW”
document_metadata.category A classification that indicates whether the source of a policy document is legislative, executive or a submission to the UNFCCC. For example, Hungary's National Inventory Report 2022 to the UNFCCC will be classified as a document 'submitted under the UNFCCC'.
document_metadata.metadata Contains the following subfields:
- framework: If the main document in the family is a framework law on climate change, this field indicates what kind of framework (i.e. mitigation or adaptation) it relates to.
- response: whether the documents in the family relate to adaptation, mitigation, loss and damage, disaster risk management or a combination of the above.
- hazard: The natural hazards referred to in a family of documents
- sector: the most relevant sectors of the global economy to which a policy document relates.
- keyword: the main concepts or themes addressed in a policy document
- instrument: Instruments are labels that provide a broad-level indication of the tools and methods used by a national government to implement particular policies and legislation. The labels have been curated by a pre-defined set of CPR policy instrument taxonomies.

UNFCCC-only:
- author: The national government or organizational body that has authored a submission to the UNFCCC.
- author_type: A classification that indicates whether the author of the policy document is a Party or Non-Party to the UNFCCC.
document_metadata.languages The primary language(s) used in a particular document
document_metadata.geographies The jurisdiction(s) whose national government or policy-making body authored the policy documents in a family. Specified as a three-letter ISO code.

(This broad definition is intended to capture supranational bodies (e.g. EU) and de facto states whose political status is contested (e.g. Taiwan)
document_metadata.translated Our pipeline translates all documents that aren’t in English to English, using the Google Cloud translation API.

This is a boolean flag which signals whether the pipeline has translated the text of the document.
document_cdn_object The address where the source document PDF can be found on CPR’s servers.
Prepend with https://cdn.climatepolicyradar.org/navigator/ to access the PDF via its URL.
document_md5_sum An md5 checksum of the data. Use if you want to check that the PDF you’ve downloaded from the source URL is the same as the one we’ve processed in the dataset.
document_content_type The content type of the original document as retrieved by the pipeline via its source URL.
If None this means we haven’t retrieved a document that’s processable by our pipeline, and there will be no text for the document.
application/pdf, text/html, None
document_metadata.family_import_id The internal system ID used by our database developers to uniquely identify a family
document_metadata.family_slug The unique identifier for this family used in the family URL. Also know as a ‘slug’. adaptation-strategy-to-climate-change-in-the-czech-republic_3c9c
You can rebuild the family URL from this ID using the format /document/<Family ID). eg. https://app.climatepolicyradar.org/document/adaptation-strategy-to-climate-change-in-the-czech-republic_3c9c
document_metadata.family_title The name used to describe the document(s) of a particular policy. In practice, the title of the main policy document will be the family name. For example, the 2020 Korean New Deal comprises the New Deal document itself and a press release about it, which together make up the family.
document_metadata.family_summary (coming soon) A short explanation of the policy contained within a family of documents. For example, a summary of Kenya’s National Adaptation Plan reads as a “document that identifies Kenya's vulnerabilities to the effect of climate change, adaptation actions and implementation strategies.”
document_metadata.corpus_import_id The internal system ID of the corpus (dataset) the document belongs to. ‘CCLW.corpus.i00000001.n0000’
document_metadata.corpus_type_name The name of the corpus (dataset) the document belongs to. ‘Laws and Policies’
document_metadata.collection_title The title of an overarching policy that unites a grouping of related documents For example, the UK’s Climate Change Act constitute a collection of policy documents, one of which include the UK’s Sixth Carbon Budget. The Budget is part of this collection because its publication is mandated under the Act.
document_metadata.collection_summary A short explanation of the overarching policy that unites a collection of related documents. For example, a summary of South Korea’s 2050 Carbon Neutral Scenario Roadmaps would look like: “The presidential committee on carbon neutrality released in October 2021 two roadmaps.They consist of two scenarios in which the domestic net emissions reach to zero in 2050. In plan A, emissions are reduced as much as possible, such as cessation of thermal power generation, while in plan B, removal technologies such as CCS are used to compensate for the remaining of some thermal power capacity.”
pipeline_metadata.parser_metadata Metadata about the version of the Azure PDF parser that was used to extract text from the document, and the date it was parsed.
document_metadata.document_role (coming soon) The role a document plays in its legislative or policy-making process (e.g. is it the main or supporting document, or an amendment)
document_metadata.document_variant (coming soon) Whether the document is written in its original language or is a transition
document_metadata.family_publication_ts The date on which a family document was published

Text block and page related

Text blocks are passages detected by the PDF or HTML parser. Text on a page or HTML document is segmented into text blocks by its layout, and segmentation isn’t necessarily the best for downstream ML applications. All text block columns are prefixed with text_block .

PDFs are structured in pages. We also store the page number of a text block and the dimensions of the page, which can be used to locate a text block on the page via its coordinates.

Column Definition Sample Data
text_block.index The index of a text block within a document. Text blocks sorted by their index are in reading order according to the Azure PDF parser.
text_block.text The text within the text block.
text_block.text_block_id A unique identifier for a text block in a document.
text_block.language The language of the text block as detected by the python langdetect library. This differs from document_metadata.languages which is a human-annotated field.
text_block.type The type of the text block as detected by the parsing method. text, sectionHeading, TableCell.
text_block.type_confidence The confidence that the parser has that the text block is of the predicted type (text_block.type).
text_block.page_number The page that the text block is on. Starts from 0.
pdf_data_page_metadata.dimensions The dimensions of the page.

Licensing

This dataset is licensed as CC by 4.0.

Please read our Terms of Use, including any specific terms relevant to commercial use. Contact [email protected] with any questions.

Useful links

Climate Policy Radar

Climate Change Laws of the World

Citation

Text citation:

Bibtex:

@misc {climate_policy_radar_2024,
    author       = { {Climate Policy Radar and Grantham Research Institute at the London School of Economics} },
    title        = { All Document Text Data },
    year         = 2024,
    url          = { https://huggingface.co/datasets/ClimatePolicyRadar/all-document-text-data },
    publisher    = { Hugging Face }
}
Downloads last month
85