GAIA/GARDIAN-CIGI: 65,550 Agricultural Research Publications

65,550 agricultural research publications form a machine-readable corpus curated by the CGIAR's GAIA project. The documents are sourced from across CGIAR centers, indexed through the GARDIAN platform, and converted from PDF to structured JSON via a pipeline using GROBID. The dataset was last updated on June 13, 2026.

Use Cases

• Train or fine-tune large language models for agricultural science based on the corpus of research publications.
• Develop document classification systems for agricultural research topics based on the indexed publication metadata.
• Build question-answering systems for agricultural knowledge based on the structured text content.
• Analyze research trends and collaboration patterns across CGIAR centers based on the indexed corpus.
• Extract structured information (e.g., methods, findings) from agricultural PDFs using the underlying GROBID pipeline.

Strengths

• Corpus contains 65,550 individual agricultural research publications.
• Documents are converted from PDF to structured JSON via a dedicated pipeline (GAIA-CIGI) using GROBID, suggesting a standardized format.
• Publications are drawn from across the CGIAR centers, indicating a multi-institutional scope.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific modeling tasks.
• Freshness should be verified as the last update timestamp is in the future (2026-06-13).

Provenance

Source

CGIAR centers, indexed through the GARDIAN platform.

Collection Method

Documents converted from PDF to structured JSON via the GAIA-CIGI pipeline using GROBID and supporting extractors.

Freshness

Last updated 2026-06-13 12:10:23; freshness should be verified.