esgTransformer: Deep Learning Models for ESG Analytics and Social Response

In this project, we will continue upgrading esgBERT based on enriched datasets and additional ESG indexes, for example, S&P Global ESG Scores. An unified ESG performance evaluation model to be introduced in this phase, which is available to analyze companies' disclosures and evaluate ESG performance among the competitors to Applied Material by using different ESG issues in interest. In addition, to support the ESG department to work more efficiently on information disclosure on social networks, this project also builds a tool based on Event Trigger and Post Generator, which offers timely social response capability. The core technology in this phase is GPT-3, designed for text generation and uses decoder architecture. With an encoder BERT, we named this project as esgTransformer.