NLP Techniques to Achieve the Detection of Climate Change in Text Corpora

Abstract

Financial intermediaries have been pressing corporations to disclose financial risks associated with climate change, particularly from individual and institutional investors. To detect these kinds of hazards in their financial and non-financial reports, companies should be required to publish a significant quantity of textual data shortly. This is especially true in light of the expanding regulations that are being enacted on the subject. To do this, this research uses cutting-edge natural language processing algorithms to identify changes in climate in text datasets. Two transformer models BERT and ClimateBert, a freshly released DistillRoBERTa-based model especially designed for climate text classification are refined using transfer learning. These two algorithms can learn contextual linkages between words in a text since they are based on the transformer architecture. To fine-tune these models, the researcher employ the novel “ClimaText” database, which contains information sourced from Wikipedia, and 10,000 file reports, including web-based claims.