Datasets:

sciosystems
/

test

---

license: cc-by-4.0
pretty_name: "GAIA / GARDIAN-CIGI Agricultural Research Corpus"
tags:
  - "science"
  - "agriculture"
  - "academic"
  - "rag"
  - "cgiar"
  - "multilingual"
  - "documents"
  - "research"
  - "food-security"
  - "climate-adaptation"
  - "sdg 1 - no poverty"
  - "nutrition"
  - "food security"
  - "africa south of sahara"
  - "carbon"
  - "climate change"
  - "latin america and the caribbean"
  - "nutrition, health and food security"
  - "poverty reduction, livelihoods and jobs"
  - "sdg 2 - zero hunger"
  - "sdg 8 - decent work and economic growth"
  - "systems transformation"
  - "decision making"
  - "health"
  - "africa"
task_categories:
  - summarization
  - text-retrieval
  - question-answering
  - text-generation
  - feature-extraction
size_categories:
  - n<1K
configs:
  - config_name: full_text_en
    default: true
    data_files:
      - split: train
        path: "FULL_TEXT/EN/Parquet/*.parquet"
  - config_name: abstracts_only_en
    data_files:
      - split: train
        path: "ABSTRACTS_ONLY/EN/Parquet/*.parquet"
  - config_name: abstracts_only_es
    data_files:
      - split: train
        path: "ABSTRACTS_ONLY/ES/Parquet/*.parquet"
  - config_name: abstracts_only_fr
    data_files:
      - split: train
        path: "ABSTRACTS_ONLY/FR/Parquet/*.parquet"
---


# GAIA / GARDIAN-CIGI Agricultural Research Corpus

This dataset contains **20** agricultural research records from the GARDIAN repository, processed through the CIGI pipeline: **6** with full text and **14** with metadata and abstract only.

## Repository Layout

```

FULL_TEXT/<LANG>/        documents with extracted full text

ABSTRACTS_ONLY/<LANG>/   documents with metadata and abstract only

    JSON/                one JSON file per document (metadata + content)

    Markdown/            one Markdown file per document (metadata front matter + text)

    Parquet/             all documents of the folder in Parquet (part files)

```

Each `FULL_TEXT`/`ABSTRACTS_ONLY` + language folder is available as a separate config in the dataset viewer, e.g. `load_dataset("sciosystems/test", "full_text_en")`.

| Folder | Documents | Tokens | Files | Size |
|---|---:|---:|---:|---:|
| `ABSTRACTS_ONLY/EN` | 9 | 0 | 20 | 146.30 KB |
| `ABSTRACTS_ONLY/ES` | 3 | 0 | 7 | 73.58 KB |
| `ABSTRACTS_ONLY/FR` | 2 | 0 | 5 | 40.12 KB |
| `FULL_TEXT/EN` | 6 | 29,954 | 13 | 468.00 KB |

## Dataset Overview

| Property | Value |
|----------|-------|
| **Total Documents** | 20 |
| **Full-Text Documents** | 6 |
| **Abstract-Only Documents** | 14 |
| **Total Size (all formats)** | 728.00 KB |
| **Total Tokens (full text)** | 29,954 |
| **Total Characters (full text)** | 188,904 |
| **Avg. Tokens per Full-Text Document** | 4,992 |
| **Total Pages (partly est.)** | 76 |
| **Languages** | 3 |
| **Countries** | 18 |
| **Unique Keywords** | 211 |
| **Resource Types** | 6 |
| **Documents with Source URL** | 20 (100.0%) |
| **Date Generated** | 2026-11-04 18:27:07 |

Tokens are counted on the `content` field of full-text documents using the `o200k_base` tokenizer (tiktoken). 3 of 6 full-text documents have a real page count; the rest are estimated at 500 words per page.

## Language Distribution

| Language | Count | Percentage |
|----------|------:|-----------:|
| English | 15 | 75.0% |
| Spanish | 3 | 15.0% |
| French | 2 | 10.0% |

## Resource Types

| Type | Count | Percentage |
|------|------:|-----------:|
| Dataset / Tabular | 7 | 35.0% |
| Scientific Publication | 6 | 30.0% |
| Report | 2 | 10.0% |
| Brief | 2 | 10.0% |
| Training Material | 2 | 10.0% |
| Dataset / Geospatial | 1 | 5.0% |

## Top Countries

| Country | Count | Percentage |
|---------|------:|-----------:|
| Peru | 3 | 15.0% |
| Viet Nam | 2 | 10.0% |
| Indonesia | 2 | 10.0% |
| Mali | 2 | 10.0% |
| India | 1 | 5.0% |
| Ethiopia | 1 | 5.0% |
| Bangladesh | 1 | 5.0% |
| Guyana | 1 | 5.0% |
| Russia | 1 | 5.0% |
| Colombia | 1 | 5.0% |
| Democratic Republic of the Congo | 1 | 5.0% |
| Malawi | 1 | 5.0% |
| Uganda | 1 | 5.0% |
| Honduras | 1 | 5.0% |
| Guatemala | 1 | 5.0% |
| Kenya | 1 | 5.0% |
| Costa Rica | 1 | 5.0% |
| Nigeria | 1 | 5.0% |

## Top Keywords

sdg 1 - no poverty, nutrition, food security, africa south of sahara, carbon, climate change, latin america and the caribbean, nutrition, health and food security, poverty reduction, livelihoods and jobs, sdg 2 - zero hunger, sdg 8 - decent work and economic growth, systems transformation, decision making, health, africa, research, emission, models, tropical forests, biomass

## Publication Years

2026 (1), 2025 (3), 2024 (2), 2023 (1), 2022 (2), 2021 (1), 2020 (1), 2018 (2), 2017 (2), 2015 (3), 2014 (2)

## Rights

Per-document licensing varies. Distribution of the `rights` field:

| Rights | Docs | Share |
|---|---:|---:|
| CC-BY | 20 | 100.0% |

> **Rights are heterogeneous.** The repo-level `cc-by-4.0` tag reflects the curation of the collection, **not** a uniform document license. **Always check the per-document `rights` field (and the source at `url`, where available) before redistributing or using individual documents commercially.**

## Document JSON Schema

```json

{

  "gardian_id": "string",

  "source": "string",

  "url": "string",

  "description": "string",

  "id": "string",

  "title": "string",

  "language": "string",

  "resource_type": "string",

  "release_year": "string",

  "rights": "string",

  "dataNODE_id": "string",

  "keywords": "list[string]",

  "geography": {

    "regions": "list[string]",

    "countries": "list[string]"

  },

  "content": "string"

}

```

## License

This dataset is released under [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/).

## Acknowledgements

This dataset was developed for the [Generative AI for Agriculture (GAIA)](https://www.ifpri.org/project/generative-ai-for-agriculture-gaia/) project, funded by the Gates Foundation and UK International Development (FCDO), in collaboration between [CGIAR](https://www.cgiar.org/) and [SCiO](https://scio.systems/).

---

*Generated automatically on 2026-11-04 18:27:07*