---
language:
- bm
- fr
- en
task_categories:
- translation
pretty_name: Bambara MT v2
size_categories:
- 100K<n<1M
tags:
- bambara
- machine-translation
---

# bambara-mt-v2

An aggregated Bambara (Bamanankan, `bm` / `bam_Latn`) machine-translation corpus pairing
Bambara with French and English, assembled from eight upstream sources.

## Load

```python
from datasets import load_dataset

# aligned table with provenance
mt = load_dataset("djelia/bambara-mt-v2", "default", split="train")

# directional training pairs
pairs = load_dataset("djelia/bambara-mt-v2", "source_target_style", split="train")
```

## Configs

| Config | Rows | Splits | Shape |
| --- | ---: | --- | --- |
| `default` | 98,086 | `train` | one row per sentence |
| `source_target_style` | 196,124 | `train`, `validation`, `test` | one row per translation direction |

Both configs hold the same corpus. Use `default` for the aligned table, `source_target_style`
for directional training pairs.

## Fields

`default`:

| Field | Description |
| --- | --- |
| `bm` | Bambara sentence |
| `en` | English translation, or an empty string |
| `fr` | French translation, or an empty string |
| `source` | Upstream slice the row came from |

`source_target_style`: `source_lang`, `target_lang`, `source_text`, `target_text`, with
ISO-639-3 + script codes (`bam_Latn`, `fra_Latn`, `eng_Latn`).

## Notes

Each `default` row carries exactly one of `en` or `fr`, never both; the unused column holds an
empty string rather than null.

`source` values: `bayelemabaga` 46,976, `lafand` 27,104, `guerre_des_griots` 6,194,
`bamadaba` 4,854, `egafe` 4,702, `google_smol_gatitos` 4,000, `google_smol_smldoc` 3,393,
`google_smolsent_en_bam` 863.

22,341 of the 98,086 `default` rows are exact duplicates of another row, almost all in the
`lafand` slice. Deduplicate before training or before computing any corpus statistic.

For evaluation, use
[`djelia/bambara-mt-benchmark`](https://huggingface.co/datasets/djelia/bambara-mt-benchmark).