README.md

metadata

language:
  - fr
  - mos
license: other
task_categories:
  - translation
pretty_name: French–Mooré (Mossi) Parallel Corpus
size_categories:
  - 1M<n<10M
tags:
  - moore
  - mossi
  - mooré
  - french
  - parallel-corpus
  - low-resource
  - glosbe
  - burkina-faso
configs:
  - config_name: default
    data_files:
      - split: train
        path: fr-mos-validated.parquet

French → Mooré (Mossi) Parallel Corpus

Machine-translated parallel sentences from French (fr) to Mooré / Mossi (mos), produced by a public-web crawl + filtering + Glosbe translation pipeline.

Snapshot

Field	Value
Validated pairs	~2.82 million
Source language	French
Target language	Mooré (Mossi)
Translator	Glosbe public MT
Export date	2026-08-03

Schema

Column	Type	Description
id	string (UUID)	Pair identifier
source_domain	string	Crawl source domain
source_url	string	Original page URL
crawl_date	string (ISO 8601)	When the page was crawled
french	string	Source sentence
moore	string	Mooré translation
translator	string	Provider id (e.g. glosbe)
translation_confidence	float	Provider confidence
complexity_score	float	Source complexity score
quality_score	float	Pipeline quality score

Files

• fr-mos-validated.jsonl — one JSON object per line
• fr-mos-validated.parquet — same rows, ZSTD-compressed Parquet

Quality notes

• Only accepted / validated pairs are included (failed validation rows are excluded).
• Translations are automatic (Glosbe); this is not a fully human-reviewed gold set.
• Source text was filtered for short/medium simple French sentences; exact duplicates were deduplicated.
• Soft confidence floor was applied during generation; average confidence on accepted pairs is roughly ~0.63.
• Dominant crawl sources include French Wikipedia and other public French sites; review licenses/terms for your use case.

Intended use

• Pretraining / fine-tuning low-resource MT models for French ↔ Mooré
• Research on West African language technology
• Bootstrapping human post-editing workflows

Limitations

• Automatic MT noise and occasional domain skew (encyclopedia, news, institutional)
• Not a substitute for native-speaker verified data for high-stakes applications
• Source provenance is retained in source_url / source_domain for auditability

Citation

If you use this dataset, please cite the dataset page and note that translations were generated with Glosbe MT via an automated corpus pipeline.