This repository contains the training data and a notebook demonstrating our model training. Our training data is a json file including about 50k text sequences annotated with 5 classes (4 TCFD categories + "None" for non-climate-related text) and the company the text was retrieved from.
If you find any error, please open an issue.
Note: This repository currently does NOT contain data or code for our new research paper "Cheap Talk in Corporate Climate Commitments: The Role of Active Institutional Ownership, Signaling, Materiality, and Sentiment". This will be uploaded soon.