Model card for climateBUG-LM

Model Description

climateBUG-LM is a deep learning language model fine-tuned for analyzing bank reports in the context of climate change and sustainability. It leverages a unique annotated corpus, climateBUG-Data, which consists of statements from EU banks' annual and sustainability reports, focusing on climate change and finance. This model aims to classify statements as relevant or irrelevant to climate-related subjects, offering enhanced performance due to its domain-specific training.

Access and Usage

Models, dataset and tools are available at the climateBUG project page.
Suitable for researchers and professionals in finance, sustainability, and climate policy.

Applications

The model is ideal for:

Analyzing financial reports for climate change-related content.
Research in financial sustainability and climate economics.
Tracking how banks articulate their climate-related activities.

Example Usage

Here is an example of how to use the climateBUG-LM model for classifying text as climate-related or not:

from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline

tokenizer_name = "lumilogic/climateBUG-LM"
model_name = "lumilogic/climateBUG-LM"


model = AutoModelForSequenceClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(tokenizer_name, max_len=512)

pipe = pipeline("text-classification", model=model,
                tokenizer=tokenizer, device_map='auto')


# Climate related text
text = 'This issue represents around 10% of the outstanding volume of green sovereign bonds and will used to finance Germany’s climate and environmental strategy.'
output = pipe(text)
print(output)  # [{'label': 'LABEL_1', 'score': 0.9974282383918762}]

# Non-climate related text
text = 'Our model, based on a customer-centric universal banking relationship, therefore demonstrated its resilience and usefulness for all stakeholders in all our regions.'
output = pipe(text)
print(output)  # [{'label': 'LABEL_0', 'score': 0.9931207299232483}]

Limitations

Optimized for EU bank reports; performance may vary for other regions.
Primarily focused on climate and finance domains.

Citation

Please cite this model as follows:

Yu, Y., Scheidegger, S., Elliott, J., & Löfgren, Å. (2024). climateBUG: A data-driven framework for analyzing bank reporting through a climate lens. Expert Systems With Applications, 239, 122162.

@article{yu2024climatebug,
title = {climateBUG : A data-driven framework for analyzing bank reporting through a climate lens},
journal = {Expert Systems with Applications},
volume = {239},
pages = {122162},
year = {2024},
author = {Yinan Yu and Samuel Scheidegger and Jasmine Elliott and Åsa Löfgren}
}

Support and Contact

For support, additional information, or inquiries, please reach out through [email protected] or visit the climateBUG project page.