climaparl-xlm-roberta-climate-change-narrow-v03

Climate classification model

This model is trained to classify climate change statements from parliamentary speeches. We finetuned a multilingual xml-roberta-base model on 7107 sentences from parliamentary speeches from 27 countries (EU + UK) to detect explicit mentions of climate change or climate change policy.

Code Book

We define sentences as climate related if they directly references climate change, climate policy, or consequences of climate change (e.g., excessive temperature rises, sea level rise). This includes explicit mentions of climate change, global warming, greenhouse gases, carbon emissions, and specific climate change policies or targets.

Examples

• "We must reduce our greenhouse gas emissions by 40% by 2030 to combat climate change."
• "We oppose the new carbon tax proposed by the government."
• "Climate change is the defining issue of our time, and we must act now to prevent catastrophic impacts."
• "The new legislation focuses on cutting carbon emissions."

Non-examples

• "We need to invest in renewable energy for a cleaner future." (Unless it explicitly mentions climate change or related policies)

Model Details

• Finetuned from model: FacebookAI/xml-roberta-base
• Learning rate: 2e-05
• Epochs: 28
• Warmup ratio: 0.1104518378674452