adzcai/babylm-eng-deu-50-50-stratified

---
language:
- eng
- deu
license: cc-by-4.0
---

# BabyLM English–German 50/50 Stratified

A bilingual training corpus for the [Multilingual BabyLM](https://github.com/adzcai/multilingual) project, combining 50% of the [English](https://huggingface.co/datasets/BabyLM-community/babylm-eng) and [German](https://huggingface.co/datasets/BabyLM-community/babylm-deu) BabyBabelLM corpora.

Sampling is **stratified by `category`**: each source category is sampled independently to preserve the original category proportions within each language.

## Token counts

| Language | Tokens | Share |
|----------|-------:|------:|
| English (`eng`) | 49,481,353 | 47.8% |
| German (`deu`) | 53,979,025 | 52.2% |
| **Total** | **103,460,378** | **100%** |

## Category breakdown

| Category | English tokens | English % | German tokens | German % |
|----------|--------------:|---------:|-------------:|--------:|
| `child-available-speech` | 4,551,767 | 9.2% | 634,123 | 1.2% |
| `child-books` | 13,406,594 | 27.1% | 1,067,431 | 2.0% |
| `child-directed-speech` | 13,857,314 | 28.0% | 3,625,926 | 6.7% |
| `child-news` | 0 | 0.0% | 1,440,298 | 2.7% |
| `child-wiki` | 7,304,856 | 14.8% | 1,340,607 | 2.5% |
| `educational` | 0 | 0.0% | 128,802 | 0.2% |
| `padding-opensubtitles` | 9,856,922 | 19.9% | 45,741,838 | 84.7% |
| `padding-wikipedia` | 503,900 | 1.0% | 0 | 0.0% |

## Source datasets

- [`BabyLM-community/babylm-eng`](https://huggingface.co/datasets/BabyLM-community/babylm-eng)
- [`BabyLM-community/babylm-deu`](https://huggingface.co/datasets/BabyLM-community/babylm-deu)