| Sumario: | This repository, developed under the CGIAR Sustainable Farming Program (SFP), establishes an AI-assisted workflow for evidence synthesis and meta-analysis in agriculture. The workflow automates labor-intensive stages of systematic review, including search term generation, screening, tagging, keyword extraction, classification, and harmonization, while preserving methodological rigor. Outputs are systematically benchmarked against expert manual reviews to quantify accuracy and relevance, providing clear guidance on where automation excels and where human oversight remains essential.
Accompanying R scripts and analytical reports ensure transparent, reproducible methods through public vignettes and generate shareable, machine-readable artifacts for downstream use. By accelerating the mechanical components of evidence synthesis, this workflow enables researchers to concentrate on interpretation, uncertainty quantification, and domain-specific questions, offering a replicable pathway to scale literature reviews and deliver decision-ready evidence.
This work has been supported by the CGIAR Climate Action Lever and represents a collaboration between the Alliance of Bioversity International and CIAT.
|
|---|