metadata

language:
  - en
  - fr
  - es
  - de
license: apache-2.0
size_categories:
  - 100K<n<1M
task_categories:
  - text-classification
pretty_name: International Aid Transparency Initiative (IATI) Policy Marker Dataset
dataset_info:
  features:
    - name: iati_identifier
      dtype: string
    - name: reporting_org_ref
      dtype: string
    - name: text
      dtype: string
    - name: languages
      dtype: string
    - name: activity_dates
      dtype: string
    - name: gender_equality_sig
      dtype: float64
    - name: environment_sig
      dtype: float64
    - name: pdgg_sig
      dtype: float64
    - name: trade_sig
      dtype: float64
    - name: bio_diversity_sig
      dtype: float64
    - name: climate_mitigation_sig
      dtype: float64
    - name: climate_adaptation_sig
      dtype: float64
    - name: desertification_sig
      dtype: float64
    - name: rmnch_sig
      dtype: float64
    - name: drr_sig
      dtype: int64
    - name: disability_sig
      dtype: int64
    - name: nutrition_sig
      dtype: int64
    - name: gender_equality
      dtype: bool
    - name: environment
      dtype: bool
    - name: pdgg
      dtype: bool
    - name: trade
      dtype: bool
    - name: bio_diversity
      dtype: bool
    - name: climate_mitigation
      dtype: bool
    - name: climate_adaptation
      dtype: bool
    - name: desertification
      dtype: bool
    - name: rmnch
      dtype: bool
    - name: drr
      dtype: bool
    - name: disability
      dtype: bool
    - name: nutrition
      dtype: bool
  splits:
    - name: train
      num_bytes: 847221626
      num_examples: 895519
  download_size: 300347056
  dataset_size: 847221626
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
tags:
  - finance
  - climate

International Aid Transparency Initiative (IATI) Policy Marker Dataset

A multi-purpose dataset including all activity title and description text published to IATI with metadata for policy markers. For more information on IATI policy markers, see the element page on the IATI Standard Website. IATI is a living data source, and this dataset was last updated on 21 August, 2024. For the code to generate an updated version of this dataset, please see my Github repository here.

For any given policy marker in this dataset, there are two columns. One indicates whether the publisher utilizes the policy marker at all (e.g. gender_equality), and the other represents the marker significance (e.g. gender_equality_sig). The language column is pipe separated. It should be mostly ISO 639, but it's a free-text field so there may be other values.

A suggested use would be:

from datasets import load_dataset

iati = load_dataset("devinitorg/iati-policy-markers", split="train")
gender_relevant_iati = iati.filter(lambda example: example["gender_equality"]).rename_column("gender_equality_sig", "label")
cols_to_remove = gender_relevant_iati.column_names
cols_to_remove.remove("text")
cols_to_remove.remove("label")
gender_relevant_iati = gender_relevant_iati.remove_columns(cols_to_remove)
dataset = gender_relevant_iati.class_encode_column("label").train_test_split(
    test_size=0.2,
    stratify_by_column="label",
    shuffle=True,
)