v3.0 update (add French, edit Tamazight) (#5)

- add fra_Latn (673f945782378f6438cd06a58495b632651ec41f)
- Tamazight corrections (2476cd0fc3dd6efe6f4333fbe4acf839697d2127)
- per-language splits, language tags in the dataset card, bib (429f870872e0a1ad273b4a28517717897751e3f9)
- Merge branch 'main' into pr/5 (ffaaf749b6fa04e3533be11fea4d4eeb2e5165c6)

CHANGELOG.md

@@ -8,8 +8,11 @@ In particular:
 1. Any updates to pre-existing files shall entail a major version bump.
 2. The inclusion of a new language, without changes to other pre-existing data, shall only entail a minor version bump.
 
-## [3.0] - 2025-11-03
-- Added `kyr_Cyrl`
+## [3.0] - 2025-11-07
+- Participants to the [WMT25 Shared Task](https://www2.statmt.org/wmt25/open-data.html) contributed multiple imprivements. Many thanks to all the contributors! Please don't forget to cite the relevant papers listed in [bibliography.bib](bibliography.bib).
+    - Added `fra_Latn` (see [Marmonier et al, 2025](https://aclanthology.org/2025.wmt-1.80/))
+    - Added `kyr_Cyrl`  (see [Jumashev et al, 2025](https://aclanthology.org/2025.wmt-1.84/))
+    - Updated about half of `zgh_Tfng` texts (see [Oktem et al, 2025](https://aclanthology.org/2025.wmt-1.82/))
 
 ## [2.0] - 2024-11-08
 

README.md

@@ -8,11 +8,234 @@ pretty_name: OLDI-Seed
 task_categories:
   - text-generation
   - translation
+language:
+  - ace
+  - ary
+  - arz
+  - bam
+  - ban
+  - ben
+  - bho
+  - bjn
+  - bug
+  - crh
+  - dik
+  - dzo
+  - eng
+  - fra
+  - fur
+  - fuv
+  - gug
+  - hne
+  - ita
+  - kas
+  - knc
+  - lij
+  - lim
+  - lmo
+  - ltg
+  - mag
+  - mni
+  - mri
+  - nqo
+  - nus
+  - pbt
+  - prs
+  - scn
+  - shn
+  - spa
+  - srd
+  - szl
+  - taq
+  - vec
+  - zgh
 configs:
   - config_name: default
     data_files:
       - split: train
         path: seed/*.parquet
+  - config_name: ace_Arab
+    data_files:
+      - split: train
+        path: seed/ace_Arab.parquet
+  - config_name: ace_Latn
+    data_files:
+      - split: train
+        path: seed/ace_Latn.parquet
+  - config_name: ary_Arab
+    data_files:
+      - split: train
+        path: seed/ary_Arab.parquet
+  - config_name: arz_Arab
+    data_files:
+      - split: train
+        path: seed/arz_Arab.parquet
+  - config_name: bam_Latn
+    data_files:
+      - split: train
+        path: seed/bam_Latn.parquet
+  - config_name: ban_Latn
+    data_files:
+      - split: train
+        path: seed/ban_Latn.parquet
+  - config_name: ben_Beng
+    data_files:
+      - split: train
+        path: seed/ben_Beng.parquet
+  - config_name: bho_Deva
+    data_files:
+      - split: train
+        path: seed/bho_Deva.parquet
+  - config_name: bjn_Arab
+    data_files:
+      - split: train
+        path: seed/bjn_Arab.parquet
+  - config_name: bjn_Latn
+    data_files:
+      - split: train
+        path: seed/bjn_Latn.parquet
+  - config_name: bug_Latn
+    data_files:
+      - split: train
+        path: seed/bug_Latn.parquet
+  - config_name: crh_Latn
+    data_files:
+      - split: train
+        path: seed/crh_Latn.parquet
+  - config_name: dik_Latn
+    data_files:
+      - split: train
+        path: seed/dik_Latn.parquet
+  - config_name: dzo_Tibt
+    data_files:
+      - split: train
+        path: seed/dzo_Tibt.parquet
+  - config_name: eng_Latn
+    data_files:
+      - split: train
+        path: seed/eng_Latn.parquet
+  - config_name: fra_Latn
+    data_files:
+      - split: train
+        path: seed/fra_Latn.parquet
+  - config_name: fur_Latn
+    data_files:
+      - split: train
+        path: seed/fur_Latn.parquet
+  - config_name: fuv_Latn
+    data_files:
+      - split: train
+        path: seed/fuv_Latn.parquet
+  - config_name: gug_Latn
+    data_files:
+      - split: train
+        path: seed/gug_Latn.parquet
+  - config_name: hne_Deva
+    data_files:
+      - split: train
+        path: seed/hne_Deva.parquet
+  - config_name: ita_Latn
+    data_files:
+      - split: train
+        path: seed/ita_Latn.parquet
+  - config_name: kas_Arab
+    data_files:
+      - split: train
+        path: seed/kas_Arab.parquet
+  - config_name: kas_Deva
+    data_files:
+      - split: train
+        path: seed/kas_Deva.parquet
+  - config_name: knc_Arab
+    data_files:
+      - split: train
+        path: seed/knc_Arab.parquet
+  - config_name: knc_Latn
+    data_files:
+      - split: train
+        path: seed/knc_Latn.parquet
+  - config_name: lij_Latn
+    data_files:
+      - split: train
+        path: seed/lij_Latn.parquet
+  - config_name: lim_Latn
+    data_files:
+      - split: train
+        path: seed/lim_Latn.parquet
+  - config_name: lmo_Latn
+    data_files:
+      - split: train
+        path: seed/lmo_Latn.parquet
+  - config_name: ltg_Latn
+    data_files:
+      - split: train
+        path: seed/ltg_Latn.parquet
+  - config_name: mag_Deva
+    data_files:
+      - split: train
+        path: seed/mag_Deva.parquet
+  - config_name: mni_Beng
+    data_files:
+      - split: train
+        path: seed/mni_Beng.parquet
+  - config_name: mri_Latn
+    data_files:
+      - split: train
+        path: seed/mri_Latn.parquet
+  - config_name: nqo_Nkoo
+    data_files:
+      - split: train
+        path: seed/nqo_Nkoo.parquet
+  - config_name: nus_Latn
+    data_files:
+      - split: train
+        path: seed/nus_Latn.parquet
+  - config_name: pbt_Arab
+    data_files:
+      - split: train
+        path: seed/pbt_Arab.parquet
+  - config_name: prs_Arab
+    data_files:
+      - split: train
+        path: seed/prs_Arab.parquet
+  - config_name: scn_Latn
+    data_files:
+      - split: train
+        path: seed/scn_Latn.parquet
+  - config_name: shn_Mymr
+    data_files:
+      - split: train
+        path: seed/shn_Mymr.parquet
+  - config_name: spa_Latn
+    data_files:
+      - split: train
+        path: seed/spa_Latn.parquet
+  - config_name: srd_Latn
+    data_files:
+      - split: train
+        path: seed/srd_Latn.parquet
+  - config_name: szl_Latn
+    data_files:
+      - split: train
+        path: seed/szl_Latn.parquet
+  - config_name: taq_Latn
+    data_files:
+      - split: train
+        path: seed/taq_Latn.parquet
+  - config_name: taq_Tfng
+    data_files:
+      - split: train
+        path: seed/taq_Tfng.parquet
+  - config_name: vec_Latn
+    data_files:
+      - split: train
+        path: seed/vec_Latn.parquet
+  - config_name: zgh_Tfng
+    data_files:
+      - split: train
+        path: seed/zgh_Tfng.parquet
+---
+Click to add a cell.
 ---
 # OLDI Seed Machine Translation Datacard
 
@@ -129,6 +352,7 @@ See [CHANGELOG.md](CHANGELOG.md) for information about the latest changes.
 | `dik` | `Latn` | `sout2832` |         Southwestern Dinka        |                                                            |
 | `dzo` | `Tibt` | `dzon1239` |              Dzongkha             |                                                            |
 | `eng` | `Latn` | `stan1293` |              English              |                                                            |
+| `fra` | `Latn` | `stan1290` |               French              |                                                            |
 | `fur` | `Latn` | `east2271` |              Friulian             |                                                            |
 | `fuv` | `Latn` | `nige1253` |         Nigerian Fulfulde         |                                                            |
 | `gug` | `Latn` | `para1311` |         Paraguayan Guaraní        |                                                            |

bibliography.bib

@@ -82,6 +82,26 @@
     publisher = "Association for Computational Linguistics"
 }
 
+% for French
+@inproceedings{marmonier-etal-2025-french,
+    title = "A {F}rench Version of the {OLDI} Seed Corpus",
+    author = "Marmonier, Malik  and
+      Sagot, Beno{\^i}t  and
+      Bawden, Rachel",
+    editor = "Haddow, Barry  and
+      Kocmi, Tom  and
+      Koehn, Philipp  and
+      Monz, Christof",
+    booktitle = "Proceedings of the Tenth Conference on Machine Translation",
+    month = nov,
+    year = "2025",
+    address = "Suzhou, China",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2025.wmt-1.80/",
+    pages = "1048--1060",
+    ISBN = "979-8-89176-341-8",
+    abstract = "We present the first French partition of the OLDI Seed Corpus, our submission to the WMT 2025 Open Language Data Initiative (OLDI) shared task. We detail its creation process, which involved using multiple machine translation systems and a custom-built interface for post-editing by qualified native speakers. We also highlight the unique translation challenges presented by the source data, which combines highly technical, encyclopedic terminology with the stylistic irregularities characteristic of user-generated content taken from Wikipedia. This French corpus is not an end in itself, but is intended as a crucial pivot resource to facilitate the collection of parallel corpora for the under-resourced regional languages of France."
+}
 % for Kyrgyz
 @inproceedings{jumashev-etal-2025-kyrgyz,
     title = "The {K}yrgyz Seed Dataset Submission to the {WMT}25 Open Language Data Initiative Shared Task",
@@ -106,4 +126,27 @@
     pages = "1088--1102",
     ISBN = "979-8-89176-341-8",
     abstract = "We present a Kyrgyz language seed dataset as part of our contribution to the WMT25 Open Language Data Initiative (OLDI) shared task. This paper details the process of collecting and curating English{--}Kyrgyz translations, highlighting the main challenges encountered in translating into a morphologically rich, low-resource language. We demonstrate the quality of the dataset through fine-tuning experiments, showing consistent improvements in machine translation performance across multiple models. Comparisons with bilingual and MNMT Kyrgyz-English baselines reveal that, for some models, our dataset enables performance surpassing pretrained baselines in both English{--}Kyrgyz and Kyrgyz{--}English translation directions. These results validate the dataset{'}s utility and suggest that it can serve as a valuable resource for the Kyrgyz MT community and other related low-resource languages."
-}
+}
+
+% for Tamazight corrections
+@inproceedings{oktem-etal-2025-correcting,
+    title = "Correcting the Tamazight Portions of {FLORES}+ and {OLDI} Seed Datasets",
+    author = "Oktem, Alp  and
+      Farhi, Mohamed Aymane  and
+      Essaidi, Brahim  and
+      Jabouja, Naceur  and
+      Boudichat, Farida",
+    editor = "Haddow, Barry  and
+      Kocmi, Tom  and
+      Koehn, Philipp  and
+      Monz, Christof",
+    booktitle = "Proceedings of the Tenth Conference on Machine Translation",
+    month = nov,
+    year = "2025",
+    address = "Suzhou, China",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2025.wmt-1.82/",
+    pages = "1072--1080",
+    ISBN = "979-8-89176-341-8",
+    abstract = "We present the manual correction of the Tamazight portions of the FLORES+ and OLDI Seed datasets to improve the quality of open machine translation resources for the language. These widely used reference corpora contained numerous issues, including mistranslations, orthographic inconsistencies, overuse of loanwords, and non-standard transliterations. Overall, 36{\%} of FLORES+ and 40{\%} of Seed sentences were corrected by expert linguists, with average token divergence of 19{\%} and 25{\%} among changed items. Evaluation of multiple MT systems, including NLLB models and commercial LLM services, showed consistent gains in automated evaluation metrics when using the corrected data. Fine-tuning NLLB-600M on the revised Seed corpus yielded improvements of +6.05 chrF (en{\textrightarrow}zgh) and +2.32 (zgh{\textrightarrow}en), outperforming larger parameter models and LLM providers in en{\textrightarrow}zgh direction."
+}

dataset_cards/fra_Latn.md

@@ -0,0 +1,58 @@
+# Dataset card
+
+## Description
+
+French partition of the OLDI Seed Corpus, consisting of segments sourced from Wikipedia articles.
+
+## License
+
+CC BY-SA 4.0
+
+## Attribution
+
+```bibtex
+@inproceedings{marmonier-etal-2025-french,
+    title = "A {F}rench Version of the {OLDI} Seed Corpus",
+    author = "Marmonier, Malik  and
+      Sagot, Beno{\^i}t  and
+      Bawden, Rachel",
+    editor = "Haddow, Barry  and
+      Kocmi, Tom  and
+      Koehn, Philipp  and
+      Monz, Christof",
+    booktitle = "Proceedings of the Tenth Conference on Machine Translation",
+    month = nov,
+    year = "2025",
+    address = "Suzhou, China",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2025.wmt-1.80/",
+    pages = "1048--1060",
+    ISBN = "979-8-89176-341-8",
+    abstract = "We present the first French partition of the OLDI Seed Corpus, our submission to the WMT 2025 Open Language Data Initiative (OLDI) shared task. We detail its creation process, which involved using multiple machine translation systems and a custom-built interface for post-editing by qualified native speakers. We also highlight the unique translation challenges presented by the source data, which combines highly technical, encyclopedic terminology with the stylistic irregularities characteristic of user-generated content taken from Wikipedia. This French corpus is not an end in itself, but is intended as a crucial pivot resource to facilitate the collection of parallel corpora for the under-resourced regional languages of France."
+}
+```
+See the paper "A French Version of the OLDI Seed Corpus" at https://aclanthology.org/2025.wmt-1.80. 
+
+## Language codes
+
+* ISO 639-3: fra
+* ISO 15924: Latn
+* Glottocode: stan1290
+
+## Additional language information
+
+The data follows the grammatical and lexical norms of standard French, as spoken and written in France. However, for consistency and to align with common practices in related datasets, it adopts a few specific typographical conventions:
+* Dactylographic apostrophes (') are used systematically instead of typographic apostrophes (’).
+* The use of French _guillemets_ (« ») with non-breaking spaces is not strictly enforced, a convention also followed in the French partitions of the FLORES dataset.
+
+For authoritative guidance on the French language, the following resources are recommended. For grammar and usage, the primary reference is M. Grevisse and A. Goosse, _Le Bon Usage_ (16th ed., De Boeck Supérieur, 2016). For a comprehensive reference in English, see H. N. Labeau & P. Larrivée, _The Cambridge Grammar of French_ (Cambridge University Press, 2022). For lexicographical aspects, the main monolingual references are _Le Grand Robert de la langue française_ and A. Rey (ed.), _Dictionnaire historique de la langue française_. A standard bilingual reference is the _Collins-Robert French Dictionary_. A recent and comprehensive reference work in English covering the French language in all its facets is W. Ayres-Bennett & M. McLaughlin (eds.), _The Oxford Handbook of the French Language_ (Oxford University Press, 2024). A high-quality open-source spell and grammar checker for French is available via [Grammalecte](https://grammalecte.net/).
+
+## Workflow
+
+This dataset was created using a machine translation post-editing (MTPE) approach. Starting from the original English partition of the OLDI Seed Corpus, we generated nine translation hypotheses per segment using a mix of systems: four "traditional" sequence-to-sequence Transformer MT models (OPUS-MT, NLLB-3.3B, NLLB-600M, MADLAD-400-3B) and five hypotheses from large language models (four from Llama 4 Scout using varied prompting strategies, and one from DeepSeek-R1).
+
+This diverse set of nine translation candidates for each segment was then presented to post-editors in a custom-built interface. The interface allowed post-editors to select any of the candidate translations, which would then populate a text area for further refinement. This refinement process focused on two key aspects of the target text: improving the fluency of the translations, a non-trivial task due to the many stylistic irregularities and errors found in the user-generated source segments originating from Wikipedia; and meticulously verifying and correcting, through extensive external research, the translation of technical terminology across the wide range of specialized, encyclopedic domains represented in the corpus.
+
+As a final step, all post-edited segments underwent a validation pass using the Grammalecte grammar checker interface.
+
+This post-editing and validation work was performed by two native French speakers with a C2 level in English. A native British English speaker with a C2 level in French was also consulted to clarify specific aspects of the source segments. Domain experts were occasionally consulted to solve terminological difficulties in highly technical domains.

dataset_cards/zgh_Tfng.md

@@ -1,15 +1,50 @@
 # Dataset card
 
-## Description
+## Description  
+Seed data in Standard Moroccan Tamazight with 2025 corrections.
 
-Seed data in Standard Moroccan Tamazight
+## License  
+CC-BY-SA-4.0  
 
-## License
+## Attribution  
+```
+@inproceedings{oktem-etal-2025-correcting,
+    title = "Correcting the Tamazight Portions of {FLORES}+ and {OLDI} Seed Datasets",
+    author = "Oktem, Alp  and
+      Farhi, Mohamed Aymane  and
+      Essaidi, Brahim  and
+      Jabouja, Naceur  and
+      Boudichat, Farida",
+    editor = "Haddow, Barry  and
+      Kocmi, Tom  and
+      Koehn, Philipp  and
+      Monz, Christof",
+    booktitle = "Proceedings of the Tenth Conference on Machine Translation",
+    month = nov,
+    year = "2025",
+    address = "Suzhou, China",
+    publisher = "Association for Computational Linguistics",
+    url = "https://aclanthology.org/2025.wmt-1.82/",
+    pages = "1072--1080",
+    ISBN = "979-8-89176-341-8",
+    abstract = "We present the manual correction of the Tamazight portions of the FLORES+ and OLDI Seed datasets to improve the quality of open machine translation resources for the language. These widely used reference corpora contained numerous issues, including mistranslations, orthographic inconsistencies, overuse of loanwords, and non-standard transliterations. Overall, 36{\%} of FLORES+ and 40{\%} of Seed sentences were corrected by expert linguists, with average token divergence of 19{\%} and 25{\%} among changed items. Evaluation of multiple MT systems, including NLLB models and commercial LLM services, showed consistent gains in automated evaluation metrics when using the corrected data. Fine-tuning NLLB-600M on the revised Seed corpus yielded improvements of +6.05 chrF (en{\textrightarrow}zgh) and +2.32 (zgh{\textrightarrow}en), outperforming larger parameter models and LLM providers in en{\textrightarrow}zgh direction."
+}
+```
 
-CC-BY-SA-4.0
+Awal initiative:
 
-## Attribution
+```
+@inproceedings{oktem-boudichat-2025-awal,
+  author    = {\"Oktem, Alp and Boudichat, Farida},
+  title     = {Awal -- Community-Powered Language Technology for Tamazight},
+  booktitle = {Proceedings of the Conf{\'e}rence Internationale sur les Technologies d'Information et de Communication pour l'Amazighe (TICAM)},
+  year      = {2025},
+  address   = {Rabat, Morocco},
+  publisher = {Institut Royal de la Culture Amazighe (IRCAM)}
+}
+```
 
+Original dataset:
 ```bibtex
 @inproceedings{seed-23,
     title = {Small Data, Big Impact: Leveraging Minimal Data for Effective Machine Translation},
@@ -23,18 +58,29 @@ CC-BY-SA-4.0
 }
 ```
 
-## Language codes
-
+## Language codes  
 * ISO 639-3: zgh
 * ISO 15924: Tfng
 * Glottocode: stan1324
 
-## Additional language information
+## Additional language information  
+Standard Moroccan Tamazight as standardized by the Royal Institute of Amazigh Culture (IRCAM), using Tifinaghe-IRCAM (Neo-Tifinagh) script. Reference materials used include:
+
+* Dictionnaire Général de la Langue Amazighe (Ameur et al., 2016)
+* Arabic-Amazigh-French Landforms Dictionary (Akioud et al., 2022)  
+* Arabic-Amazigh Dictionary (Chafik, 1996)
+* Phonologie de l'Amazighe (Boukous, 2009)
+* La Nouvelle Grammaire de l'Amazighe (Boukhris et al., 2008)
+* Grammaire Générative de l'Amazighe: Morphologie et Syntaxe du Nom (El Moujahid, 2022)
+* Manuel de Conjugaison de l'Amazighe (Laabdelaoui et al., 2012)
+
+## Workflow  
 
-Reference dictionary: IRCAM’s [Dictionnaire Général de la Langue Amazighe Informatisé](https://tal.ircam.ma/dglai/lexieam.php).
+The original data was released as part of the NLLB-Seed dataset, where it was incorrectly labeled `tzm_Tfng`. It was relabeled as `zgh_Tfng` after community feedback and additional quality assessment. Please refer to the paper for further information.
 
-## Workflow
+The 2025 update submission contains manual corrections of Standard Moroccan Tamazight (zgh) translations that addressed mistranslations, orthographic inconsistencies, overuse of loanwords, and non-standard transliterations in the original datasets. The corrected data includes 997 sentences from FLORES+ dev, 1,012 sentences from FLORES+ devtest, and 6,193 sentences from OLDI Seed, with 36% of FLORES+ and 40% of Seed sentences requiring corrections.
 
-This data was released as part of the NLLB-Seed dataset, where it was incorrectly labeled `tzm_Tfng`. It was relabeled as `zgh_Tfng` after community feedback and additional quality assessment. Please refer to the paper for further information.
+Corrections were performed by expert linguists taking English source sentences as reference. The FLORES+ dev and devtest sets were revised in full in two iterations by two linguists. The OLDI Seed dataset was divided into batches of 1,000 sentences and distributed among three professional Tamazight translators. Quality control was ensured through spot-checking of each linguist's work by another linguist through random sampling. All translators were professional translators and were compensated as part of this work.
 
-## Additional guidelines
+## Additional guidelines  
+Corrections followed IRCAM standardization guidelines, prioritizing native Tamazight terms over loanwords when available. For loanwords that were retained, proper Tamazight morphological patterns were applied (e.g., appropriate nominal prefixes). Transliterations of foreign words and names followed standardized practices, with particular attention to accurate mapping of sounds not present in Tamazight to their closest equivalents.

seed/fra_Latn.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d2a1cc90e5d6823fa90724a1488f779e1a4f0c132c24d9eb57ccedb49401c0c
+size 684394

seed/zgh_Tfng.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:61815ce14b41049e930117f73f8393ae0664a731be2a0f14edd347d6fa6d0ee7
-size 876509
+oid sha256:fccffbc6b9a35215d658cf93674d80965030fd088ebe3df6daa2dd1ff6424dc8
+size 875585