Biodiversity Commitment Classifier Model Overview: This binary text classification model identifies and distinguishes biodiversity commitments in corporate sustainability reports. It classifies paragraphs as either:

Commitment: The company pledges specific actions to improve biodiversity or reduce negative environmental impacts Non-commitment: General statements, observations, or non-actionable content related to biodiversity

Model Architecture: Built on ClimateBERT, a DistilRoBERTa-based model pre-trained on climate-related text, this classifier was fine-tuned specifically for biodiversity commitment detection in corporate disclosures. Training Data: The model was trained on a curated dataset of 2,000 manually annotated paragraphs extracted from sustainability reports of Fortune Global 500 companies, ensuring high-quality labels and real-world applicability to corporate ESG analysis.

Recommended Pipeline: First use ESGBERT/EnvironmentalBERT-biodiversity to identify biodiversity-related paragraphs, then apply this model to identify commitments.

Performance Metrics: Average of 5-fold cross-validation: Weighted F10.928 Weighted Precision0.930Weighted Recall0.929AUC-ROC0.976

Downloads last month
9
Safetensors
Model size
82.3M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support