A taxonomy creation for agriculture using classical machine learning algorithms

Abstract

The Ceres2030 is an evidence and cost modeling program to support donor-decision making on high-impact interventions needed to end hunger and transform the lives of the world's poorest farmers (Sustainable Development Goal 2). Policy and decision-makers are interested in finding useful techniques and approaches to address urgent problems. Our goal was to automate the process of finding interventions, a colloquially used term, in articles and to make it easier for researchers and non-researchers search for scientific achievements. We used machine learning semantic models to generate a taxonomy of agricultural interventions and outcomes relevant to policy-makers. The intervention classifier was built with the help of classical machine learning algorithms, and our first results show the possibility of making use of even small datasets for natural language processing tasks.