1.1. Digital for impact: digital capabilities as accelerators to support the transformation of agrifood systems 31
1.1.4. Making sense of the data
Extracting insights from project documents through machine learning
Renewed efforts were made this year to refine the theory and practice of applying machine learning to extract insights from FAO project documents. The Organization implements roughly 2 000 projects in any one biennium. Although many documents are written to describe the purpose, objectives and desired outcomes of these projects, it is not easy to extract this information for further analysis (e.g. comparing projects in one region). The processes that support the compilation of the FAO digital portfolio (FDP) quantify projects by business and digital thematic areas, which are manually assigned. This work aims to improve the completeness and
©rawpixel.com/Freepik
efficiency of the FDP’s underlying processes by automating the assignment of these thematic areas. The work culminated in June 2022 with the completion of a graduate-level thesis in partnership with the Technical University of Denmark.
External partnerships to build capabilities
Collaboration with University of Southampton for Small Island Developing States (SIDS) FAO, through the HiH, is working with Southampton University, the United Kingdom of Great Britain and Northern Ireland, to extend the datasets of multi-hazard and vulnerability data for Pacific Island SIDS, and now includes other global SIDS, providing data and tools via the FAO HiH Geospatial Platform.
Data include high resolution (100m grid) information on geophysical, meteorological, hydrological and marine hazards for a baseline and
217 tables covering 122 islands/strata in Kiribati (2019/2020), Solomon Islands (2012), Tuvalu (2017) and Vanuatu (2019/2020). Source: Southampton University
global average temperature scenarios (1.5 °C and 2.0 °C), rather than on specific time periods. The projects are exploring approaches to develop socioagricultural vulnerability data together with exposure data at the same resolution to develop hazard-specific risk indices that can inform policies and response strategies. The projects are also providing workshops and training in the data to regional and national entities and developing plans for further training to enable the effective utilization of the data in country by national bodies, researchers and NGOs.
Made with FlippingBook - Online Brochure Maker