Workshop SDS: Registration
View Submission - SDS2022
A0180
Title: Difference in SDG reporting of research articles using zero-shot text classification Authors:  Elena Toenjes - Justus-Liebig-University Giessen (Germany)
Lutz Breuer - Justus-Liebig-University Giessen (Germany)
Ramona Teuber - Justus-Liebig-University Giessen (Germany)
Christoph Funk - Justus-Liebig-University Giessen (Germany) [presenting]
Abstract: In September 2015, the United Nations (UN) set an agenda to transform our world by 2030 with the adoption of 17 Sustainable Development Goals (SDGs), 169 targets and 231 indicators for monitoring progress towards the goals. Since then, the academic literature on the SDGs has grown tremendously. The analysis of such large amounts of textual data requires the use of Natural language processing (NLP) techniques. Here, we apply zero-shot classification as a text mining tool on SDG-related scientific articles to analyze the scientific discourse on the 17 SDGs. The contributions are four-fold. First, we review the scientific literature on the SDGs in order to draw conclusions about the scientific discourses worldwide. Second, we show that abstracts contain the most relevant information from scientific articles related to the discussed SDGs. This means that applying NLP techniques on abstracts instead of the whole article is sufficient, which in turn saves computational power and thus time. Third, we show that zero-shot text classification can be a useful tool to label extensive textual information and thus might be relevant for policymakers by providing information beyond the typical UN indicators in an efficient manner. Fourth, we compare the scientific discourse with the official average UN SDG indicator scores.