metadata
license: apache-2.0
language:
- en
- fr
- es
- it
- de
tags:
- synthetic
- RAG
- DSE
- retrieval
size_categories:
- 100K<n<1M
task_categories:
- visual-document-retrieval
- text-retrieval
WIP - there might be issues with the negatives
OGC - Organized, Grouped, Cleaned
Intended for image/text to vector (DSE)
Dataset Composition
The dataset merges, shuffles, and formats data from:
Dataset Statistics
| Metric | Value |
|---|---|
| Total rows | 700,000+ |
| Rows with negatives | ≈ 33% |
| Rows without queries (image negatives only) | ≈ 25% |
Language Distribution
| Language | Ratio |
|---|---|
| English | ≈ 52% |
| French | ≈ 12% |
| Spanish | ≈ 12% |
| Italian | ≈ 12% |
| German | ≈ 12% |
Creators
Dataset curated by:
- Paul Lemaistre
- Léo Appourchaux