Automatic ESG Assessment of Companies by Mining and Evaluating Media Coverage Data: NLP Approach and Tool

Replication package for our paper submission to ECIR 2023

It contains the following files:

1. Our data set of 487,621 news headlines annotated as being environmental-, governance-, or social-related. We encourage fellow researchers to use the corpus as a benchmark for other ESG-relevant NLP tasks.
2. Code/Notebooks that we used for the training and evaluation of our company detection, ESG classification, and sentiment models
3. Full tables detailing the results of all experiments performed