0-shot text classification for web-based environmental indicators: Pilot study on B-Corp data
[EN] This paper proposes a tool that uses web-based information to generate a proxy for the environmental culture indicator developed by B-Lab. The tool is based on recent advances in Natural Language Processing (NLP), such as pre-trained language models like BART that better capture the semantic fa...
| Autores: | , , , |
|---|---|
| Tipo de documento: | capítulo de livro |
| Data de publicação: | 2023 |
| País: | España |
| Recursos: | Universitat Politècnica de València (UPV) |
| Repositório: | RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia |
| Idioma: | inglês |
| OAI Identifier: | oai:riunet.upv.es:10251/201682 |
| Acesso em linha: | https://riunet.upv.es/handle/10251/201682 |
| Access Level: | Acceso aberto |
| Palavra-chave: | Natural Language Processing Zero-shot text classification Sustainable Innovation |
| Resumo: | [EN] This paper proposes a tool that uses web-based information to generate a proxy for the environmental culture indicator developed by B-Lab. The tool is based on recent advances in Natural Language Processing (NLP), such as pre-trained language models like BART that better capture the semantic facets of natural language. The algorithm and data provide several advantages, including real-time analysis, minimal building cost, granularity, and a large sample size, making it appealing. The Zero-shot text classification task is used to create an indicator of companies' environmental culture, which was chosen due to the urgency created by recent climatic events, pushing for increased environmental protection and sustainability culture promotion. The tool was tested on the B-CORP dataset, which provides scores on environmental performance. Results indicate that scores for certain environmental topics generated by the tool are correlated with B-Lab's environmental indicator. This research open door to the possibility of predicting the environmental readiness of the companies base on web-based indicators. |
|---|