voyager205/sdg-finetuned-enhanced

SDG Fine-tuned Enhanced Model

A Sentence Transformer model fine-tuned for SDG (Sustainable Development Goals) alignment tasks. This model is designed to classify and analyze text activities according to the 17 UN Sustainable Development Goals.

Model Description

• Base Model: all-mpnet-base-v2
• Fine-tuned for: SDG alignment and classification
• Embedding Dimension: 768
• Max Sequence Length: 384
• Language: English

Intended Use

This model is specifically trained to:

1. Align activities with SDGs: Determine which SDG(s) a given activity or text passage relates to
2. Semantic similarity: Measure how similar activities are in the context of SDG alignment
3. Text classification: Classify text into SDG categories for sustainability reporting

Primary Use Cases

• Local government annual report analysis for SDG alignment
• Corporate sustainability reporting
• Research on SDG-related activities
• Automated SDG tagging systems

Usage

Using Sentence Transformers

from sentence_transformers import SentenceTransformer

# Load the model
model = SentenceTransformer('voyager205/sdg-finetuned-enhanced')

# Encode sentences
sentences = [
    "The council implemented a new recycling program to reduce waste.",
    "New solar panels were installed on community buildings.",
    "The health department launched a vaccination campaign."
]

embeddings = model.encode(sentences)
print(f"Embedding shape: {embeddings.shape}")  # (3, 768)

For SDG Alignment

from sentence_transformers import SentenceTransformer
import numpy as np

model = SentenceTransformer('voyager205/sdg-finetuned-enhanced')

# SDG descriptions or keywords
sdg_texts = [
    "No Poverty - End poverty in all its forms everywhere",
    "Zero Hunger - End hunger, achieve food security",
    "Good Health and Well-being - Ensure healthy lives",
    # ... all 17 SDGs
]

# Encode SDG descriptions and activities
sdg_embeddings = model.encode(sdg_texts)
activity_embedding = model.encode(["The council built affordable housing units."])

# Find most aligned SDG
similarities = np.dot(activity_embedding, sdg_embeddings.T)
most_aligned_sdg = np.argmax(similarities) + 1
print(f"Most aligned SDG: {most_aligned_sdg}")

Training Details

Training Data

The model was fine-tuned on council annual reports and SDG-related activities from Australian local governments, including:

• Activity descriptions from annual reports
• SDG-aligned text samples for each of the 17 SDGs
• Manually curated and validated training examples

Training Configuration

• Base Model: all-mpnet-base-v2
• Fine-tuning Method: Contrastive learning with SDG-specific pairs
• Training Framework: Sentence Transformers

Performance

This enhanced model provides improved accuracy for SDG alignment tasks compared to the base model:

• Better semantic understanding of SDG-related content
• More accurate classification of government activities
• Improved handling of Australian local government terminology

Limitations

• Trained primarily on Australian local government documents
• May have reduced accuracy for other contexts or regions
• English language only
• Best suited for activity-level text (sentences to paragraphs)

Ethical Considerations

• This model should be used as a tool to assist human analysis, not replace it
• Results should be validated by domain experts for critical applications
• Consider potential biases in the training data

Citation

If you use this model in your research, please cite:

@software{sdg_finetuned_enhanced,
  author = {voyager205},
  title = {SDG Fine-tuned Enhanced Model},
  year = {2025},
  url = {https://huggingface.co/voyager205/sdg-finetuned-enhanced}
}

License

MIT License

Contact

For questions or issues with this model, please open an issue on the Hugging Face repository.