norbert4-base-scandinavian-embedding

NorBERT4-base Scandinavian Embedding Model

Multi-dataset trained embedding model for Norwegian, Danish, and Swedish languages.

Model Details

• Base Model: ltg/norbert4-base
• Embedding Dimension: 640
• Max Sequence Length: 256 tokens
• Languages: Norwegian (Bokmål & Nynorsk), Danish, Swedish
• Training Approach: Multi-dataset ROUND_ROBIN sampling

Training Data

Total: 1.6M samples across 3 Scandinavian languages

1. NLI Dataset (556k samples, Norwegian)

• Source: Fremtind/all-nli-norwegian
• Format: (anchor, positive, negative) triplets
• Purpose: Natural language understanding and semantic similarity

2. Question-Answering Dataset (100k samples, NO+DA+SV)

• NorQuAD: ltg/norquad - Norwegian QA
• NorBookQA: ltg/norbookqa - Norwegian OpenBookQA
• ScandiQA: alexandrainst/scandi-qa - Scandinavian QA (NO+DA+SV)
• Supervised-DA: Danish sentence pairs
• Format: (query, positive) pairs
• Purpose: Question-document retrieval

3. DDSC Nordic Dataset (949k samples, NO+DA+SV)

• Source: DDSC/nordic-embedding-training-data
• Format: (query, positive, [negative]) pairs
• Composition: 40% with hard negatives, 60% with in-batch negatives
• Purpose: General retrieval with hard negative mining

Training Details

• Strategy: ROUND_ROBIN multi-dataset sampling (prevents catastrophic forgetting)
• Batch Size: 16 (effective batch size: 32 with gradient accumulation)
• Learning Rate: 5e-6 (low LR to prevent overfitting)
• Epochs: 1 epoch through all datasets
• Loss: MultipleNegativesRankingLoss
• Early Stopping: Tracks average loss across all three datasets
• Regularization: weight_decay=0.015, no warmup

Performance

MTEB Retrieval Benchmarks (vs previous models):

Usage

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("thivy/norbert4-base-scandinavian-embedding")

# Encode sentences
sentences = [
    "Dette er en norsk setning",
    "Detta är en svensk mening", 
    "Dette er en dansk sætning"
]

embeddings = model.encode(sentences)

# Compute similarity
from sentence_transformers.util import cos_sim
similarity = cos_sim(embeddings[0], embeddings[1])

Intended Use

• Semantic search across Scandinavian languages
• Document retrieval and ranking
• Question-answering systems
• Cross-lingual similarity (NO/DA/SV)
• Text clustering and classification

Limitations

• Max sequence length: 256 tokens (longer texts are truncated)
• Optimized for Scandinavian languages (Norwegian, Danish, Swedish)
• Best performance on retrieval tasks (not instruction-following)

Citation

If you use this model, please cite:

@misc{norbert4-scandi-embedding,
  title={NorBERT4 Scandinavian Embedding Model},
  author={Thivyesh Ahilathasan},
  year={2025},
  url={https://huggingface.co/thivy/norbert4-base-scandinavian-embedding}
}

Related Models

• Base: ltg/norbert4-base
• Large: thivy/norbert4-large-scandinavian-embedding (coming soon)