Climatebug: A Data-Driven Framework for Analyzing Bank Reporting Through a Climate Lens
15 Pages Posted: 20 Dec 2022
Abstract
In this article we present a framework, climateBUG, that offers a framework to detect latent information about how banks discuss their activities related to climate change. The framework is based on a corpus of European banks’ annual reports, and it consists of an ingestion pipeline to extract information from the corpus data, a configurable database and a set of API’s. In addition, climateBUG provides two standalone components that can be used for customized analyses: 1. A unique annotated corpus with the foci of climate change and finance with approximately 1.1M (1,070,070) data points offered open access that can be used as training data for further model optimization. 2. A deep learning model adapted to the corpus. Benchmarking of the classification performance of the model shows that it outperforms other models with similar foci. The framework has been developed using an interdisciplinary approach with relevant domain knowledge from financial reporting, climate economics, and advanced computational linguistics natural language processing. The intended users of climateBUG include academic researchers, government representatives, managers from the the banking industry, and journalists.
Keywords: natural language processing, Annual reporting, climate change, sustainability, Finance & accounting
Suggested Citation: Suggested Citation