ateSDG: Automatic Extraction of Sustainable Development Goals
Politics Project

#Berlin #SDGs #Economy #PublicPolicy
January – December 2023

ateSDG: Automatic Extraction of Sustainable Development Goals

Funded institution(s):
Technische Universität Berlin
Cooperation partner(s):
German Research Center for Artificial Intelligence (DFKI)

General information

The project uses machine learning algorithms to automatically extract information about sustainable development goals (SDGs) from text documents. This will enable stakeholders to analyze which SDGs are addressed in the sustainability reports of companies and organizations in the Berlin-Brandenburg region. The work of the RENN hubs (Regional Hubs for Sustainability Strategies), which support and promote the implementation of the SDGs in their respective regions, can be supported in this way.


The first step involved creating a data bank with information on sustainability reports from companies and organizations in the Berlin-Brandenburg region and ground truth labels on SDGs. Various classifiers based on artificial neural networks were trained and tested on the SDG classification dataset. The project also investigated how the comprehensibility and transparency of the model results can be improved.


The classification shows promising results. The usefulness of explanations in the context of SDG recognition was examined and confirmed through a user study. In addition, the models trained on the data and the evaluated explanations were integrated into the interactive demonstrator.


A paper titled “A Transfer Learning Approach for SDGs Classification of Sustainability Reports” was published in Human Language Technologies as a Challenge for Computer Science and Linguistics 2023. Further publications are currently being prepared. The SDG demonstrator will be presented at the Association for Computational Linguistics 2024 and subsequently made publicly available.


Vera Schmitt


Vera Schmitt
Technische Universität Berlin