README.md

---
license: cc-by-4.0
task_categories:
- sentence-similarity
- feature-extraction
language:
- en
tags:
- biomedical
- scientific-literature
- pubmed
- pmc
- embeddings
- soda-vec
- negative-sampling
- sentence-transformers
size_categories:
- 10M<n<100M
---

# SODA-VEC Paired Dataset for Negative Sampling

This is a **paired version** of the SODA-VEC dataset, specifically formatted for negative sampling training with `MultipleNegativesRankingLoss`.

## Dataset Overview

- **Total examples**: 26,573,900
- **Format**: Paired (anchor-positive) for contrastive learning
- **Source**: [EMBO/soda-vec-data-full_pmc_title_abstract](https://huggingface.co/datasets/EMBO/soda-vec-data-full_pmc_title_abstract)
- **Purpose**: Training sentence transformers with negative sampling

## Data Format

Each example contains:
- **`anchor`** (string): The title of the scientific article
- **`positive`** (string): The abstract of the scientific article  
- **`pmcid`** (string): PubMed Central ID for reference

## Usage

```python
from datasets import load_dataset
from sentence_transformers import SentenceTransformer
from sentence_transformers.losses import MultipleNegativesRankingLoss

# Load the paired dataset
dataset = load_dataset("EMBO/soda-vec-data-full_pmc_title_abstract_paired")

# Use with MultipleNegativesRankingLoss
model = SentenceTransformer('all-MiniLM-L6-v2')
loss = MultipleNegativesRankingLoss(model)

# The dataset is ready for training
train_dataset = dataset['train']
```

## Training Example

```python
from sentence_transformers import SentenceTransformerTrainer, SentenceTransformerTrainingArguments

# Training arguments
args = SentenceTransformerTrainingArguments(
    output_dir="./soda-vec-negative-sampling",
    num_train_epochs=3,
    per_device_train_batch_size=32,
    learning_rate=2e-5,
    fp16=True,
)

# Create trainer
trainer = SentenceTransformerTrainer(
    model=model,
    args=args,
    train_dataset=dataset['train'],
    eval_dataset=dataset['validation'],
    loss=loss,
)

# Start training
trainer.train()
```

## Citation

If you use this dataset, please cite the original SODA-VEC work:

```bibtex
@software{soda_vec_2025,
  title={SODA-VEC: Optimized Sentence Transformer for Biomedical Text},
  author={EMBO},
  year={2025},
  url={https://github.com/source-data/soda-vec},
  note={Dataset: EMBO/soda-vec-data-full_pmc_title_abstract_paired}
}
```

## License

This dataset is released under the **CC-BY-4.0** license, consistent with PubMed Central's open access requirements.