Datasets:
File size: 3,178 Bytes
88fadbb bb85f7d 88fadbb bb85f7d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 | ---
license: mit
language:
- fr
size_categories:
- 1M<n<10M
task_categories:
- text-classification
- text-regression
tags:
- medical
- french
- biomedical
- clinical
- annotations
pretty_name: Biomed-FR-v3 Complete Multi-Task Annotated Dataset
---
# Biomed-FR-v3 Complete Multi-Task Annotated Dataset
This dataset contains French biomedical text annotated with **20 different classification and regression tasks** using the `rntc/biomed-fr-v2-classifier` model.
## Dataset Summary
- **Total samples**: 2,942,667
- **Total columns**: 41
- **Annotation tasks**: 25
- **Language**: French
- **Domain**: Biomedical/Clinical
- **Filter criteria**: Complete dataset with all 20 annotation tasks
## Key Features
- ✅ **Complete annotation coverage**: All 20 tasks from biomed-fr-v2-classifier
- ✅ **Includes `rewriting_needed`**: Critical regression task for content quality
- ✅ **Quality metrics**: Educational scores, terminology precision, content richness
- ✅ **Clinical focus**: Medical subfield classification, clinical case detection
- ✅ **Proper column order**: Original educational_score preserved (1-5 scale)
## Annotation Tasks
### Regression Tasks (15)
- `rewriting_needed`: Content rewriting necessity score
- `contains_bias`: Bias detection score
- `writing_quality`: Text quality assessment
- `terminology_precision`: Medical terminology accuracy
- `content_richness`: Information density score
- Plus others: age_group, assertion_type, certainty_level, etc.
### Classification Tasks (5)
- `medical_subfield`: 45 medical specialties
- `content_type`: 9 content categories
- `writing_style`: 5 writing styles
- `text_type`: meaningful vs incomplete
- `interactive_elements`: 4 interaction types
## Usage
```python
from datasets import load_dataset
# Load the dataset
dataset = load_dataset("rntc/bb-tt-3")
# Access key annotations
texts = dataset["train"]["text"]
rewriting_scores = dataset["train"]["rewriting_needed"]
educational_scores = dataset["train"]["educational_score"] # Original 1-5 scale
medical_fields = dataset["train"]["medical_subfield"]
```
## Data Quality
- All samples processed with consistent batch processing
- Original educational_score preserved (0.58-5.10 scale)
- Regression outputs clearly separated (e.g., educational_score_predicted)
- Dimension mismatches handled for classification tasks
- Complete 20-task coverage including previously missing regression tasks
## Model Information
Annotations generated using:
- **Model**: `rntc/biomed-fr-v2-classifier`
- **Base model**: `almanach/camembertv2-base`
- **Tasks**: 20 multi-task classification and regression heads
- **Key fix**: Restored original educational_score column
## Citation
```bibtex
@dataset{biomed_fr_v3_annotated,
title={Biomed-FR-v3 Complete Multi-Task Annotated Dataset},
author={RNTC Research Team},
year={2024},
url={https://huggingface.co/datasets/rntc/bb-tt-3},
note={French biomedical corpus with complete 20-task annotations}
}
```
## License
MIT License - see LICENSE file for details.
## Related Datasets
- **Pretraining subset**: `rntc/bb-tt-3-pretrain`
- **High quality subset**: `rntc/bb-tt-3-s3`, `rntc/bb-tt-3-s4`
|