v1.5.3 - 590 verified entries, provenance recovered, licence caveat stated

Synchronises the published dataset with the platform database, which had drifted.

WHAT CHANGED
- 607 -> 590 entries. 18 AI-generated entries that carried status "verified" were
withdrawn on provenance, not correctness: nothing attested them. Most are ordinary
Sango words; they were presented as attested when nothing attested them.
- Provenance recovered for 121 entries whose `source` named the import route rather than
the origin. 39 csv-import and 43 admin rows are now textbook-myc-csv and
textbook-myc-manual - both from the MYC (Kirindoni) CAR primary-school Sango primer.
Two rows created during a 2026-04-20 native-speaker audit now name that session.
- The first community-contributed entry ("yanga da" / porte) is published for the first
time; it was verified on 2026-09-02 and had never reached this dataset.

WHAT THE CARD NOW SAYS THAT IT DID NOT
- The dominant upstream host behind the NLLB-derived entries is named: of the 70 rows
carrying a source_url, 64 point at jw.org. Previously that appeared only in a
limitations note about example sentences, not in the licence section.
- The reasoning for why the entries are still usable is stated rather than implied: a
headword and its gloss are facts about a language, not authored expression. The EU sui
generis database right is flagged as an open question we have not taken advice on.
- Known defect recorded: every row reads status "verified", which means the curation
pipeline accepted it. Exactly ONE entry is confirmed by a native speaker. The
native-speaker-audit-2026-04-20 source is NOT a native-speaker confirmation and is
counted separately.

The native-speaker-confirmed count remains 1. It would have been 3 if entries created
during an audit were counted as confirmed by one; they are not.

DATASHEET.md

@@ -7,7 +7,7 @@ computed directly from the published files (v1.4.7) and are honest about what th
 
 - **Dataset**: `MEYNG/sango-vocabulary` — https://huggingface.co/datasets/MEYNG/sango-vocabulary
 - **Version**: 1.4.7 · **Maintainer**: MEYNG (`[email protected]`)
-- **Files**: `vocabulary.jsonl` (607 entries), `training_pairs.jsonl` (360 pairs), `grammar_rules.json` (13 rules)
+- **Files**: `vocabulary.jsonl` (590 entries), `training_pairs.jsonl` (360 pairs), `grammar_rules.json` (13 rules)
 - **Languages**: Sango (`sag`), French (`fra`), English (`eng`)
 - **Licence**: CC-BY-SA-4.0
 
@@ -27,26 +27,28 @@ computed directly from the published files (v1.4.7) and are honest about what th
 ## 2. Composition
 
 - **What do instances represent?** Two kinds:
-  1. **Vocabulary** (607 rows): a Sango word/phrase with French + English translations,
+  1. **Vocabulary** (590 rows): a Sango word/phrase with French + English translations,
      a semantic `category`, a `difficulty` level, optional example sentences, a
      pipeline `confidence` score (0–1, present on 100% of rows), a `status`, and a
      `source` provenance tag.
   2. **Training pairs** (360 rows): Sango→French text pairs extracted from a grammar/
      dictionary reference.
-- **How many instances?** 607 vocabulary entries; 360 translation pairs; 13 grammar
+- **How many instances?** 590 vocabulary entries; 360 translation pairs; 13 grammar
   rules. Vocabulary is **unique by `(sango, french)`**; 4 Sango spellings recur with
   distinct French senses (**polysemes**, deliberately kept — not duplicates).
 - **Is it a sample or complete?** It is the current verified subset of a larger,
   growing curation pipeline (4,995 additional candidates await native-speaker review).
 - **What is the "verified" label?** `status: verified` is a **curation-pipeline** flag
   (cross-referenced against textbook/dictionary/corpus sources). It is **not**
-  native-speaker sign-off. Provenance of the 607 entries:
+  native-speaker sign-off. Provenance of the 590 entries:
 
   | Provenance group                            | Entries |
   | ------------------------------------------- | ------- |
   | NLLB parallel-corpus enrichment             | 316     |
-  | Textbook (CAR primer / dictionary)          | 174     |
-  | Admin / CSV import                          | 88      |
+  | Textbook (CAR primer / dictionary)          | 256     |
+  | Origin not recorded                         | 6       |
+  | Native-speaker audit session (2026-04-20)   | 2       |
+  | Community contributor                       | 1       |
   | AI prompt enrichment                        | 18      |
   | Swadesh list                                | 3       |
   | Wiktionary                                  | 1       |
@@ -56,7 +58,7 @@ computed directly from the published files (v1.4.7) and are honest about what th
 - **Are there labels/targets?** The French/English translations are the targets for MT;
   `category`/`difficulty` support classification and curriculum tasks.
 - **Missing information?** Example sentences cover ~84.8% (Sango) / ~68.1% (French);
-  English examples absent (504 of 607 carry a Sango example after the 2026-07-22 removal of 18 that did not contain their headword). **No pronunciation/audio field** (0/607) — a known gap for a
+  English examples absent (482 of 590 carry a Sango example). **No pronunciation/audio field** (0/590) — a known gap for a
   tonal language. Some `training_pairs.jsonl` rows are fragments, not clean sentences.
 - **Confidential / sensitive / personal data?** None. The data is general-domain
   vocabulary; it contains no personal, identifying, or sensitive information.

README.md

@@ -62,7 +62,7 @@ This dataset provides trilingual vocabulary entries (Sango-French-English) with
 
 ### Vocabulary (`vocabulary.jsonl`)
 
-607 verified trilingual vocabulary entries (v1.4.9), each containing:
+590 verified trilingual vocabulary entries (v1.5.0), each containing:
 
 | Field            | Type   | Description                                                      | Coverage |
 | ---------------- | ------ | ---------------------------------------------------------------- | -------- |
@@ -146,16 +146,18 @@ The vocabulary spans 36 semantic categories. In v1.4.8 the nine singular/plural
 
 Vocabulary entries were structured and categorized by MEYNG's founder, a native Sango speaker, drawing on textbook, dictionary, and parallel-corpus sources. Each entry carries a `confidence` score and a `status` flag from the curation pipeline. **Most entries are pipeline-verified (cross-referenced against the sources above) rather than individually confirmed by a native speaker** — broader native-speaker re-verification, via a community panel, is in progress.
 
-### Provenance breakdown (v1.4.8, all 607 entries)
+### Provenance breakdown (v1.5.0, all 590 entries)
 
-`status: verified` is a _pipeline_ flag. The table below shows how each entry actually entered the set, so users can judge reliability directly rather than trusting a single "verified" label. **Only 1 of 607 entries has so far been individually confirmed by a native speaker** (`source: native_speaker`); the rest are textbook-, corpus-, or AI-derived and cross-referenced. Growing the native-speaker-confirmed share is the dataset's active roadmap (see the SangoAI vocabulary-verification workflow).
+`status: verified` is a _pipeline_ flag. The table below shows how each entry actually entered the set, so users can judge reliability directly rather than trusting a single "verified" label. **Only 1 of 590 entries has so far been individually confirmed by a native speaker** (`source: native_speaker`); the rest are textbook-, corpus-, or AI-derived and cross-referenced. Growing the native-speaker-confirmed share is the dataset's active roadmap (see the SangoAI vocabulary-verification workflow).
 
 | Provenance group                            | Entries | Origin                                                         |
 | ------------------------------------------- | ------- | -------------------------------------------------------------- |
 | NLLB parallel-corpus enrichment             | 316     | `nllb-enrichment`, `nllb_corpus`, `nllb_en_corpus`             |
-| Textbook (CAR primer / dictionary)          | 174     | `textbook-myc-review`, `textbook-batch-import`, `textbook-myc` |
-| Admin / CSV import                          | 88      | `admin`, `csv-import`                                          |
-| AI prompt enrichment                        | 18      | `prompt-vocabulary`                                            |
+| Textbook (CAR primer / dictionary)          | 256     | `textbook-myc-review`, `textbook-batch-import`, `textbook-myc`, `textbook-myc-csv`, `textbook-myc-manual` |
+| Origin not recorded                         | 6       | `admin`                                                        |
+| Native-speaker audit session (2026-04-20)   | 2       | `native-speaker-audit-2026-04-20` — **not** native-*confirmed*, see below |
+| Community contributor                       | 1       | `contributor`                                                  |
+| AI prompt enrichment                        | **0**   | `prompt-vocabulary` — **all 18 withdrawn 2026-09-04**          |
 | Swadesh list                                | 3       | `swadesh_list`                                                 |
 | Wiktionary                                  | 1       | `fr_wiktionary`                                                |
 | **Native speaker (individually confirmed)** | **1**   | `native_speaker`                                               |
@@ -167,10 +169,10 @@ For the full question-by-question provenance and collection detail, see [`DATASH
 
 ### Known Limitations
 
-1. **Dataset size**: This release contains **607 verified vocabulary entries** (v1.4.8) and **360 translation pairs**, plus `grammar_rules.json` with 13 structured Sango grammar rules. The vocabulary was consolidated and deduplicated from a larger raw set (capitalization variants and wrong-spelling collisions removed). All 607 entries are unique by `(sango, french)`; **1** Sango spelling (`mû`) recurs with distinct French senses — a genuine polyseme, glossed both "prendre, saisir" and "donner" with different argument structures (`mû X` vs `mû X na Y`). **Corrected 2026-07-22:** v1.4.0 stated 4 such spellings and called them all legitimate polysemes. Three were not — `yongoro` (same sense reworded), `dê` (a "winter" gloss lifted from 2 Timothy 4:21, not a Sango sense) and `séléka` (a "mariage" gloss; Séléka means *alliance* in Sango). All three were duplicates from a double corpus-import on 2026-05-21 and have been removed.
+1. **Dataset size**: This release contains **590 verified vocabulary entries** (v1.5.0) and **360 translation pairs**, plus `grammar_rules.json` with 13 structured Sango grammar rules. The vocabulary was consolidated and deduplicated from a larger raw set (capitalization variants and wrong-spelling collisions removed). All 590 entries are unique by `(sango, french)`; **1** Sango spelling (`mû`) recurs with distinct French senses — a genuine polyseme, glossed both "prendre, saisir" and "donner" with different argument structures (`mû X` vs `mû X na Y`). **Corrected 2026-07-22:** v1.4.0 stated 4 such spellings and called them all legitimate polysemes. Three were not — `yongoro` (same sense reworded), `dê` (a "winter" gloss lifted from 2 Timothy 4:21, not a Sango sense) and `séléka` (a "mariage" gloss; Séléka means *alliance* in Sango). All three were duplicates from a double corpus-import on 2026-05-21 and have been removed.
 2. **Verification semantics**: `status: verified` is a curation-pipeline flag (cross-referenced against textbook, dictionary, and parallel-corpus sources). The majority of entries were machine-enriched or textbook-imported; only a small fraction have been individually confirmed by a native speaker so far. Treat `confidence` accordingly and apply your own filtering for high-stakes use.
 3. **Training pair quality**: The translation pairs contain some fragmented text from the source material. Filtering is recommended for tasks requiring clean parallel data.
-4. **Example sentences**: Roughly 483 of 607 entries (~79.6%) include a Sango example sentence and ~398 (~65.6%) include a French example; English examples are not yet included. **Coverage fell again in v1.4.9, also on purpose:** 21 example sentences were removed because their provenance is the jw.org-dominated NLLB bitext (1,289 of 1,406 entries carrying a `source_url` point at that host) and one — for `apendere` — was recognisably Acts 2:11, i.e. text from a copyrighted Bible *translation* that cannot be relicensed under CC-BY-SA. The other 20 were short and almost certainly carried no protected expression, but shared the provenance and the register; a general-purpose lexicon whose examples skew religious is less useful for the health, education and administrative domains this dataset is meant to serve. Two of the removed sentences were additionally **not Sango** — `"Kristo azali Yesu."` and `"Testament ya Yesu."` use *azali* and *ya*, which are Lingala forms (Sango: *ayeke*, *ti*); this was spotted by tooling, not by a speaker, and awaits confirmation. The 21 headwords and their glosses are unchanged and remain published; only the example field was cleared, and all 21 are queued for native-speaker elicitation. **Coverage had already fallen from 84.8% in v1.4.1 on purpose:** 18 example sentences that did not contain their own headword were removed on 2026-07-22 rather than left in place — `séléka` was illustrated by a sentence not containing the word `séléka`, `doli` (*éléphant*) by one about a monkey. They are **not** being replaced with generated text: writing plausible Sango to fill the gap would inject invented language into a corpus whose value is that it documents a real one. Attested replacement candidates have been mined from the aligned corpus and await native-speaker approval. Every remaining example now contains its headword, apart from two multi-word entries matched discontinuously (correct Sango).
+4. **Example sentences**: 482 of 590 entries (81.7%) include a Sango example sentence and 397 (67.3%) include a French example; English examples are not yet included. **Coverage fell again in v1.4.9, also on purpose:** 21 example sentences were removed because their provenance is the jw.org-dominated NLLB bitext (1,289 of 1,406 entries carrying a `source_url` point at that host) and one — for `apendere` — was recognisably Acts 2:11, i.e. text from a copyrighted Bible *translation* that cannot be relicensed under CC-BY-SA. The other 20 were short and almost certainly carried no protected expression, but shared the provenance and the register; a general-purpose lexicon whose examples skew religious is less useful for the health, education and administrative domains this dataset is meant to serve. Two of the removed sentences were additionally **not Sango** — `"Kristo azali Yesu."` and `"Testament ya Yesu."` use *azali* and *ya*, which are Lingala forms (Sango: *ayeke*, *ti*); this was spotted by tooling, not by a speaker, and awaits confirmation. The 21 headwords and their glosses are unchanged and remain published; only the example field was cleared, and all 21 are queued for native-speaker elicitation. **Coverage had already fallen from 84.8% in v1.4.1 on purpose:** 18 example sentences that did not contain their own headword were removed on 2026-07-22 rather than left in place — `séléka` was illustrated by a sentence not containing the word `séléka`, `doli` (*éléphant*) by one about a monkey. They are **not** being replaced with generated text: writing plausible Sango to fill the gap would inject invented language into a corpus whose value is that it documents a real one. Attested replacement candidates have been mined from the aligned corpus and await native-speaker approval. Every remaining example now contains its headword, apart from two multi-word entries matched discontinuously (correct Sango).
 5. **Pronunciation**: The current release does **not** include a pronunciation field. Learner-oriented phonetic guides are available in the SangoAI app and are planned for a future dataset version.
 6. **Dialect coverage**: Sango has regional variations. This dataset primarily represents the standardized Sango used in Bangui (the capital).
 
@@ -218,7 +220,7 @@ This dataset is **curated**, not wholly original. The CC-BY-SA-4.0 grant covers
 
 ### Source attribution (required in addition to the above)
 
-**316 of the 607 vocabulary entries derive from the NLLB web-mined Sango-French parallel corpus**, via quality-filtered extraction (LASER >= 1.0, target-language-ID >= 0.9). That corpus is distributed by AllenAI under **ODC-BY**, which requires attribution. If you redistribute or adapt this dataset, carry that credit forward alongside MEYNG's:
+**316 of the 590 vocabulary entries derive from the NLLB web-mined Sango-French parallel corpus**, via quality-filtered extraction (LASER >= 1.0, target-language-ID >= 0.9). That corpus is distributed by AllenAI under **ODC-BY**, which requires attribution. If you redistribute or adapt this dataset, carry that credit forward alongside MEYNG's:
 
 > Contains material derived from the NLLB web-mined bitext (AllenAI distribution,
 > ODC-BY). NLLB Team et al., *No Language Left Behind: Scaling Human-Centered
@@ -231,7 +233,50 @@ Two clarifications, because these are routinely confused:
 
 The AllenAI card also states that users are *"bound to the respective Terms of Use and License of the original source"* -- the NLLB bitext is web-mined, so individual sentences carry whatever terms their origin site imposes. ODC-BY governs the compilation, not every underlying page.
 
-A further 174 entries derive from a Central African Republic language primer / dictionary reference. That work has not been identified precisely enough to state its terms or write an attribution for it; we are recording the gap rather than implying it does not exist.
+A further 256 entries derive from a Central African Republic language primer / dictionary reference (the MYC / *Kîrîndönî* primary-school Sango primer). That work has not been identified precisely enough to state its terms or write an attribution for it; we are recording the gap rather than implying it does not exist.
+
+### Where the material actually comes from
+
+Counted directly from the published `vocabulary.jsonl` at v1.5.0 (2026-09-04):
+
+| Provenance (`source` field) | Entries |
+| --- | --- |
+| NLLB web-mined bitext (`nllb-enrichment`, `nllb_corpus`, `nllb_en_corpus`) | 316 |
+| Named non-NLLB sources (CAR primer 256, Swadesh 3, audit session 2, French Wiktionary 1, native speaker 1, community contributor 1) | 264 |
+| `admin` — origin not recorded beyond the import route | 6 |
+| `prompt-vocabulary` — AI-prompt sourced | **0** (all 18 withdrawn 2026-09-04) |
+| no `source` value | 6 |
+
+**The dominant upstream host is jw.org.** Of the 70 entries carrying a `source_url`, **64 point there** — a single publisher whose own terms do not permit redistribution. (A larger figure, 1,289 of 1,406, appears under Limitations above; that describes the *mining pool* entries were selected from, not these 590 published rows. They are different denominators and should not be quoted interchangeably.)
+
+We do not think this makes the vocabulary entries unusable, and the reasoning belongs in the open rather than left implied:
+
+- **A headword and its gloss are facts about a language, not authored expression.** The 316 NLLB-derived entries are lexical facts extracted from aligned sentences.
+- **Example sentences are a different matter**, because a sentence can carry protected expression. That is why 21 were removed in v1.4.9 — one was recognisably Acts 2:11 from a copyrighted Bible translation — and why replacements are elicited from native speakers rather than mined.
+
+**This is our reasoning, not legal advice, and one question is genuinely open:** the EU *sui generis* database right protects substantial extraction from a database independently of whether its contents are facts. We have not taken advice on whether it applies here.
+
+**If you need a licence chain clean to a named source**, filter `source` to the 264 entries in the "named non-NLLB" row above. The 10 `admin` / unlabelled entries are **not** a clean subset — their provenance is unrecorded or generated, which is a separate problem from the NLLB chain and is not solved by avoiding it.
+
+### `native-speaker-audit-2026-04-20` is not a native-speaker confirmation
+
+Two entries carry that source. They were **created during** a native-speaker audit session on
+2026-04-20 — a real, dated event — but that is **not** the same as having been confirmed by a
+speaker, and they are counted in their own provenance bucket rather than folded into
+`native_speaker_confirmed`.
+
+The evidence for keeping them separate: **that same audit updated 19 other entries and changed
+the source on none of them** (14 primer, 3 NLLB-derived, 2 since withdrawn). If the convention
+were "audited by a speaker ⇒ native-sourced", those 19 would say so. And the entries that *do*
+read `native_speaker` were created a month later by an unrelated process.
+
+**So the native-speaker-confirmed count remains 1.** Folding these in would have made it 3 on
+the strength of a session name.
+
+### Known defect: `status` does not carry the verification tier
+
+Every one of the 590 rows reads `status: verified`. **"Verified" here means the curation pipeline accepted the entry.** (The 18 AI-sourced entries that previously carried this flag were withdrawn on 2026-09-04 — see the provenance table.) It does not mean a native speaker confirmed it — exactly **one** entry has that, as stated under Limitations. The intended three-tier model (`candidate` / `pipeline_verified` / `native_verified`) is **not yet represented in this file**. Until it is, read `status` as "present in the curated set" and nothing stronger.
+
 
 ## Acknowledgments
 

dataset_card.md

@@ -62,7 +62,7 @@ This dataset provides trilingual vocabulary entries (Sango-French-English) with
 
 ### Vocabulary (`vocabulary.jsonl`)
 
-607 verified trilingual vocabulary entries (v1.4.9), each containing:
+590 verified trilingual vocabulary entries (v1.5.0), each containing:
 
 | Field            | Type   | Description                                                      | Coverage |
 | ---------------- | ------ | ---------------------------------------------------------------- | -------- |
@@ -146,16 +146,18 @@ The vocabulary spans 36 semantic categories. In v1.4.8 the nine singular/plural
 
 Vocabulary entries were structured and categorized by MEYNG's founder, a native Sango speaker, drawing on textbook, dictionary, and parallel-corpus sources. Each entry carries a `confidence` score and a `status` flag from the curation pipeline. **Most entries are pipeline-verified (cross-referenced against the sources above) rather than individually confirmed by a native speaker** — broader native-speaker re-verification, via a community panel, is in progress.
 
-### Provenance breakdown (v1.4.8, all 607 entries)
+### Provenance breakdown (v1.5.0, all 590 entries)
 
-`status: verified` is a _pipeline_ flag. The table below shows how each entry actually entered the set, so users can judge reliability directly rather than trusting a single "verified" label. **Only 1 of 607 entries has so far been individually confirmed by a native speaker** (`source: native_speaker`); the rest are textbook-, corpus-, or AI-derived and cross-referenced. Growing the native-speaker-confirmed share is the dataset's active roadmap (see the SangoAI vocabulary-verification workflow).
+`status: verified` is a _pipeline_ flag. The table below shows how each entry actually entered the set, so users can judge reliability directly rather than trusting a single "verified" label. **Only 1 of 590 entries has so far been individually confirmed by a native speaker** (`source: native_speaker`); the rest are textbook-, corpus-, or AI-derived and cross-referenced. Growing the native-speaker-confirmed share is the dataset's active roadmap (see the SangoAI vocabulary-verification workflow).
 
 | Provenance group                            | Entries | Origin                                                         |
 | ------------------------------------------- | ------- | -------------------------------------------------------------- |
 | NLLB parallel-corpus enrichment             | 316     | `nllb-enrichment`, `nllb_corpus`, `nllb_en_corpus`             |
-| Textbook (CAR primer / dictionary)          | 174     | `textbook-myc-review`, `textbook-batch-import`, `textbook-myc` |
-| Admin / CSV import                          | 88      | `admin`, `csv-import`                                          |
-| AI prompt enrichment                        | 18      | `prompt-vocabulary`                                            |
+| Textbook (CAR primer / dictionary)          | 256     | `textbook-myc-review`, `textbook-batch-import`, `textbook-myc`, `textbook-myc-csv`, `textbook-myc-manual` |
+| Origin not recorded                         | 6       | `admin`                                                        |
+| Native-speaker audit session (2026-04-20)   | 2       | `native-speaker-audit-2026-04-20` — **not** native-*confirmed*, see below |
+| Community contributor                       | 1       | `contributor`                                                  |
+| AI prompt enrichment                        | **0**   | `prompt-vocabulary` — **all 18 withdrawn 2026-09-04**          |
 | Swadesh list                                | 3       | `swadesh_list`                                                 |
 | Wiktionary                                  | 1       | `fr_wiktionary`                                                |
 | **Native speaker (individually confirmed)** | **1**   | `native_speaker`                                               |
@@ -167,10 +169,10 @@ For the full question-by-question provenance and collection detail, see [`DATASH
 
 ### Known Limitations
 
-1. **Dataset size**: This release contains **607 verified vocabulary entries** (v1.4.8) and **360 translation pairs**, plus `grammar_rules.json` with 13 structured Sango grammar rules. The vocabulary was consolidated and deduplicated from a larger raw set (capitalization variants and wrong-spelling collisions removed). All 607 entries are unique by `(sango, french)`; **1** Sango spelling (`mû`) recurs with distinct French senses — a genuine polyseme, glossed both "prendre, saisir" and "donner" with different argument structures (`mû X` vs `mû X na Y`). **Corrected 2026-07-22:** v1.4.0 stated 4 such spellings and called them all legitimate polysemes. Three were not — `yongoro` (same sense reworded), `dê` (a "winter" gloss lifted from 2 Timothy 4:21, not a Sango sense) and `séléka` (a "mariage" gloss; Séléka means *alliance* in Sango). All three were duplicates from a double corpus-import on 2026-05-21 and have been removed.
+1. **Dataset size**: This release contains **590 verified vocabulary entries** (v1.5.0) and **360 translation pairs**, plus `grammar_rules.json` with 13 structured Sango grammar rules. The vocabulary was consolidated and deduplicated from a larger raw set (capitalization variants and wrong-spelling collisions removed). All 590 entries are unique by `(sango, french)`; **1** Sango spelling (`mû`) recurs with distinct French senses — a genuine polyseme, glossed both "prendre, saisir" and "donner" with different argument structures (`mû X` vs `mû X na Y`). **Corrected 2026-07-22:** v1.4.0 stated 4 such spellings and called them all legitimate polysemes. Three were not — `yongoro` (same sense reworded), `dê` (a "winter" gloss lifted from 2 Timothy 4:21, not a Sango sense) and `séléka` (a "mariage" gloss; Séléka means *alliance* in Sango). All three were duplicates from a double corpus-import on 2026-05-21 and have been removed.
 2. **Verification semantics**: `status: verified` is a curation-pipeline flag (cross-referenced against textbook, dictionary, and parallel-corpus sources). The majority of entries were machine-enriched or textbook-imported; only a small fraction have been individually confirmed by a native speaker so far. Treat `confidence` accordingly and apply your own filtering for high-stakes use.
 3. **Training pair quality**: The translation pairs contain some fragmented text from the source material. Filtering is recommended for tasks requiring clean parallel data.
-4. **Example sentences**: Roughly 483 of 607 entries (~79.6%) include a Sango example sentence and ~398 (~65.6%) include a French example; English examples are not yet included. **Coverage fell again in v1.4.9, also on purpose:** 21 example sentences were removed because their provenance is the jw.org-dominated NLLB bitext (1,289 of 1,406 entries carrying a `source_url` point at that host) and one — for `apendere` — was recognisably Acts 2:11, i.e. text from a copyrighted Bible *translation* that cannot be relicensed under CC-BY-SA. The other 20 were short and almost certainly carried no protected expression, but shared the provenance and the register; a general-purpose lexicon whose examples skew religious is less useful for the health, education and administrative domains this dataset is meant to serve. Two of the removed sentences were additionally **not Sango** — `"Kristo azali Yesu."` and `"Testament ya Yesu."` use *azali* and *ya*, which are Lingala forms (Sango: *ayeke*, *ti*); this was spotted by tooling, not by a speaker, and awaits confirmation. The 21 headwords and their glosses are unchanged and remain published; only the example field was cleared, and all 21 are queued for native-speaker elicitation. **Coverage had already fallen from 84.8% in v1.4.1 on purpose:** 18 example sentences that did not contain their own headword were removed on 2026-07-22 rather than left in place — `séléka` was illustrated by a sentence not containing the word `séléka`, `doli` (*éléphant*) by one about a monkey. They are **not** being replaced with generated text: writing plausible Sango to fill the gap would inject invented language into a corpus whose value is that it documents a real one. Attested replacement candidates have been mined from the aligned corpus and await native-speaker approval. Every remaining example now contains its headword, apart from two multi-word entries matched discontinuously (correct Sango).
+4. **Example sentences**: 482 of 590 entries (81.7%) include a Sango example sentence and 397 (67.3%) include a French example; English examples are not yet included. **Coverage fell again in v1.4.9, also on purpose:** 21 example sentences were removed because their provenance is the jw.org-dominated NLLB bitext (1,289 of 1,406 entries carrying a `source_url` point at that host) and one — for `apendere` — was recognisably Acts 2:11, i.e. text from a copyrighted Bible *translation* that cannot be relicensed under CC-BY-SA. The other 20 were short and almost certainly carried no protected expression, but shared the provenance and the register; a general-purpose lexicon whose examples skew religious is less useful for the health, education and administrative domains this dataset is meant to serve. Two of the removed sentences were additionally **not Sango** — `"Kristo azali Yesu."` and `"Testament ya Yesu."` use *azali* and *ya*, which are Lingala forms (Sango: *ayeke*, *ti*); this was spotted by tooling, not by a speaker, and awaits confirmation. The 21 headwords and their glosses are unchanged and remain published; only the example field was cleared, and all 21 are queued for native-speaker elicitation. **Coverage had already fallen from 84.8% in v1.4.1 on purpose:** 18 example sentences that did not contain their own headword were removed on 2026-07-22 rather than left in place — `séléka` was illustrated by a sentence not containing the word `séléka`, `doli` (*éléphant*) by one about a monkey. They are **not** being replaced with generated text: writing plausible Sango to fill the gap would inject invented language into a corpus whose value is that it documents a real one. Attested replacement candidates have been mined from the aligned corpus and await native-speaker approval. Every remaining example now contains its headword, apart from two multi-word entries matched discontinuously (correct Sango).
 5. **Pronunciation**: The current release does **not** include a pronunciation field. Learner-oriented phonetic guides are available in the SangoAI app and are planned for a future dataset version.
 6. **Dialect coverage**: Sango has regional variations. This dataset primarily represents the standardized Sango used in Bangui (the capital).
 
@@ -218,7 +220,7 @@ This dataset is **curated**, not wholly original. The CC-BY-SA-4.0 grant covers
 
 ### Source attribution (required in addition to the above)
 
-**316 of the 607 vocabulary entries derive from the NLLB web-mined Sango-French parallel corpus**, via quality-filtered extraction (LASER >= 1.0, target-language-ID >= 0.9). That corpus is distributed by AllenAI under **ODC-BY**, which requires attribution. If you redistribute or adapt this dataset, carry that credit forward alongside MEYNG's:
+**316 of the 590 vocabulary entries derive from the NLLB web-mined Sango-French parallel corpus**, via quality-filtered extraction (LASER >= 1.0, target-language-ID >= 0.9). That corpus is distributed by AllenAI under **ODC-BY**, which requires attribution. If you redistribute or adapt this dataset, carry that credit forward alongside MEYNG's:
 
 > Contains material derived from the NLLB web-mined bitext (AllenAI distribution,
 > ODC-BY). NLLB Team et al., *No Language Left Behind: Scaling Human-Centered
@@ -231,7 +233,50 @@ Two clarifications, because these are routinely confused:
 
 The AllenAI card also states that users are *"bound to the respective Terms of Use and License of the original source"* -- the NLLB bitext is web-mined, so individual sentences carry whatever terms their origin site imposes. ODC-BY governs the compilation, not every underlying page.
 
-A further 174 entries derive from a Central African Republic language primer / dictionary reference. That work has not been identified precisely enough to state its terms or write an attribution for it; we are recording the gap rather than implying it does not exist.
+A further 256 entries derive from a Central African Republic language primer / dictionary reference (the MYC / *Kîrîndönî* primary-school Sango primer). That work has not been identified precisely enough to state its terms or write an attribution for it; we are recording the gap rather than implying it does not exist.
+
+### Where the material actually comes from
+
+Counted directly from the published `vocabulary.jsonl` at v1.5.0 (2026-09-04):
+
+| Provenance (`source` field) | Entries |
+| --- | --- |
+| NLLB web-mined bitext (`nllb-enrichment`, `nllb_corpus`, `nllb_en_corpus`) | 316 |
+| Named non-NLLB sources (CAR primer 256, Swadesh 3, audit session 2, French Wiktionary 1, native speaker 1, community contributor 1) | 264 |
+| `admin` — origin not recorded beyond the import route | 6 |
+| `prompt-vocabulary` — AI-prompt sourced | **0** (all 18 withdrawn 2026-09-04) |
+| no `source` value | 6 |
+
+**The dominant upstream host is jw.org.** Of the 70 entries carrying a `source_url`, **64 point there** — a single publisher whose own terms do not permit redistribution. (A larger figure, 1,289 of 1,406, appears under Limitations above; that describes the *mining pool* entries were selected from, not these 590 published rows. They are different denominators and should not be quoted interchangeably.)
+
+We do not think this makes the vocabulary entries unusable, and the reasoning belongs in the open rather than left implied:
+
+- **A headword and its gloss are facts about a language, not authored expression.** The 316 NLLB-derived entries are lexical facts extracted from aligned sentences.
+- **Example sentences are a different matter**, because a sentence can carry protected expression. That is why 21 were removed in v1.4.9 — one was recognisably Acts 2:11 from a copyrighted Bible translation — and why replacements are elicited from native speakers rather than mined.
+
+**This is our reasoning, not legal advice, and one question is genuinely open:** the EU *sui generis* database right protects substantial extraction from a database independently of whether its contents are facts. We have not taken advice on whether it applies here.
+
+**If you need a licence chain clean to a named source**, filter `source` to the 264 entries in the "named non-NLLB" row above. The 10 `admin` / unlabelled entries are **not** a clean subset — their provenance is unrecorded or generated, which is a separate problem from the NLLB chain and is not solved by avoiding it.
+
+### `native-speaker-audit-2026-04-20` is not a native-speaker confirmation
+
+Two entries carry that source. They were **created during** a native-speaker audit session on
+2026-04-20 — a real, dated event — but that is **not** the same as having been confirmed by a
+speaker, and they are counted in their own provenance bucket rather than folded into
+`native_speaker_confirmed`.
+
+The evidence for keeping them separate: **that same audit updated 19 other entries and changed
+the source on none of them** (14 primer, 3 NLLB-derived, 2 since withdrawn). If the convention
+were "audited by a speaker ⇒ native-sourced", those 19 would say so. And the entries that *do*
+read `native_speaker` were created a month later by an unrelated process.
+
+**So the native-speaker-confirmed count remains 1.** Folding these in would have made it 3 on
+the strength of a session name.
+
+### Known defect: `status` does not carry the verification tier
+
+Every one of the 590 rows reads `status: verified`. **"Verified" here means the curation pipeline accepted the entry.** (The 18 AI-sourced entries that previously carried this flag were withdrawn on 2026-09-04 — see the provenance table.) It does not mean a native speaker confirmed it — exactly **one** entry has that, as stated under Limitations. The intended three-tier model (`candidate` / `pipeline_verified` / `native_verified`) is **not yet represented in this file**. Until it is, read `status` as "present in the curated set" and nothing stronger.
+
 
 ## Acknowledgments
 

metadata.json

@@ -1,138 +1,140 @@
-{
-  "dataset_name": "meyng/sango-vocabulary",
-  "version": "1.4.9",
-  "creation_date": "2026-04-02",
-  "last_updated": "2026-08-08",
-  "description": "First structured digital vocabulary dataset for the Sango language (ISO 639-1: sg), the national language of the Central African Republic spoken by over 5 million people.",
-  "note": "Counts below are computed directly from vocabulary.jsonl (the published artifact): 607 entries, all status=verified, unique by (sango, french). The live SangoAI API (sangoai.sbs) reports its own verified count from DynamoDB, which grows over time and may differ from this file snapshot. 'verified' is a curation-pipeline flag, not native-speaker sign-off: only 1 of 607 entries is source=native_speaker. See DATASHEET.md for full provenance. Corrected 2026-07-22: this statistics block had not been recomputed since v1.4.1 and had drifted from the file it describes across five releases (it claimed 4 polysemes where the README said 1, 554 Sango examples where the file held 504, and 217 textbook entries where 41 had already been withdrawn). Every field here is now recomputed from vocabulary.jsonl.",
-  "languages": [
-    {
-      "code": "sg",
-      "name": "Sango",
-      "bcp47": "sg",
-      "iso639_3": "sag"
-    },
-    {
-      "code": "fr",
-      "name": "French",
-      "bcp47": "fr",
-      "iso639_3": "fra"
-    },
-    {
-      "code": "en",
-      "name": "English",
-      "bcp47": "en",
-      "iso639_3": "eng"
-    }
-  ],
-  "license": "CC-BY-SA-4.0",
-  "source": {
-    "organization": "MEYNG",
-    "website": "https://meyng.com",
-    "platform": "SangoAI (https://sangoai.sbs)",
-    "repository": "https://github.com/meyng-hub/sangoai"
-  },
-  "statistics": {
-    "vocabulary": {
-      "total_entries": 607,
-      "unique_by_sango_french": 607,
-      "unique_sango_spellings": 606,
-      "polysemes_same_sango_diff_french": 1,
-      "with_confidence_score": 435,
-      "with_example_sango": 483,
-      "with_example_french": 398,
-      "with_pronunciation_field": 0,
-      "with_pronunciation_note": "The published JSONL carries no pronunciation field (0/607). The live SangoAI database holds pronunciation guides for a subset; exposing them in the dataset is a planned future version.",
-      "categories": 36,
-      "provenance": {
-        "nllb_corpus_enrichment": 316,
-        "textbook_primer_dictionary": 174,
-        "admin_csv_import": 88,
-        "ai_prompt_enrichment": 18,
-        "swadesh_list": 3,
-        "wiktionary": 1,
-        "native_speaker_confirmed": 1,
-        "unlabelled": 6
-      },
-      "category_distribution": {
-        "essential": 162,
-        "verbs": 59,
-        "places": 54,
-        "adjectives": 30,
-        "numbers": 28,
-        "actions": 24,
-        "food": 21,
-        "time": 20,
-        "emotions": 18,
-        "family": 17,
-        "noun": 17,
-        "animals": 16,
-        "greetings": 14,
-        "verb_action": 12,
-        "body": 11,
-        "professions": 10,
-        "health": 8,
-        "technology": 8,
-        "clothing": 7,
-        "colors": 7,
-        "commerce": 7,
-        "nature": 7,
-        "object": 7,
-        "education": 5,
-        "grammar": 5,
-        "home": 5,
-        "particle": 5,
-        "questions": 5,
-        "adverb": 4,
-        "phrase": 4,
-        "transport": 3,
-        "verb_motion": 2,
-        "verb_state": 2,
-        "pronoun": 1,
-        "tool": 1,
-        "weather": 1
-      },
-      "difficulty_distribution": {
-        "beginner": 338,
-        "intermediate": 219,
-        "advanced": 95,
-        "expert": 1
-      }
-    },
-    "training_pairs": {
-      "total_pairs": 360,
-      "pair_directions": {
-        "sango_to_french": 360
-      },
-      "note": "Extracted from a Sango grammar and dictionary reference. Includes vocabulary entries, sentence fragments, and contextual translations. Quality varies; some entries are book fragments rather than clean parallel sentences."
-    }
-  },
-  "files": [
-    {
-      "filename": "vocabulary.jsonl",
-      "description": "Structured trilingual vocabulary entries (Sango-French-English), deduplicated",
-      "format": "JSONL",
-      "entries": 607
-    },
-    {
-      "filename": "grammar_rules.json",
-      "description": "Structured Sango grammar rules compiled from AI error analysis (13 rules, v1.1.0)",
-      "format": "JSON",
-      "entries": 13
-    },
-    {
-      "filename": "training_pairs.jsonl",
-      "description": "Sango-French translation pairs for fine-tuning",
-      "format": "JSONL",
-      "entries": 360
-    }
-  ],
-  "citation": {
-    "title": "Sango Vocabulary Dataset",
-    "authors": [
-      "MEYNG"
-    ],
-    "year": 2026,
-    "url": "https://huggingface.co/datasets/meyng/sango-vocabulary"
-  }
-}
+{
+  "dataset_name": "meyng/sango-vocabulary",
+  "version": "1.5.3",
+  "creation_date": "2026-04-02",
+  "last_updated": "2026-09-04",
+  "description": "First structured digital vocabulary dataset for the Sango language (ISO 639-1: sg), the national language of the Central African Republic spoken by over 5 million people.",
+  "note": "Every figure in the statistics block below is recomputed from vocabulary.jsonl - the published artifact itself, not the database - by scripts/data/regenerate_published_dataset.py. Read the counts there; this note deliberately restates none of them, because a prose copy of a number is a copy that drifts. The live SangoAI API (sangoai.sbs) reports its own verified count from the platform database, which changes over time - it is not monotonic, and falls when entries are withdrawn - so it may differ from this snapshot. 'verified' is a curation-pipeline flag, NOT native-speaker sign-off; see provenance.native_speaker_confirmed below and DATASHEET.md for the full chain. History: this block had not been recomputed since v1.4.1 and had drifted from the file it describes across five releases, corrected 2026-07-22; this note itself then drifted, still claiming 607 entries at v1.5.3, corrected 2026-09-04 by removing the restated numbers rather than updating them.",
+  "languages": [
+    {
+      "code": "sg",
+      "name": "Sango",
+      "bcp47": "sg",
+      "iso639_3": "sag"
+    },
+    {
+      "code": "fr",
+      "name": "French",
+      "bcp47": "fr",
+      "iso639_3": "fra"
+    },
+    {
+      "code": "en",
+      "name": "English",
+      "bcp47": "en",
+      "iso639_3": "eng"
+    }
+  ],
+  "license": "CC-BY-SA-4.0",
+  "source": {
+    "organization": "MEYNG",
+    "website": "https://meyng.com",
+    "platform": "SangoAI (https://sangoai.sbs)",
+    "repository": "https://github.com/meyng-hub/sangoai"
+  },
+  "statistics": {
+    "vocabulary": {
+      "total_entries": 590,
+      "unique_by_sango_french": 590,
+      "unique_sango_spellings": 589,
+      "polysemes_same_sango_diff_french": 1,
+      "with_confidence_score": 417,
+      "with_example_sango": 482,
+      "with_example_french": 397,
+      "with_pronunciation_field": 0,
+      "with_pronunciation_note": "The published JSONL carries no pronunciation field (0/607). The live SangoAI database holds pronunciation guides for a subset; exposing them in the dataset is a planned future version.",
+      "categories": 36,
+      "provenance": {
+        "nllb_corpus_enrichment": 316,
+        "textbook_primer_dictionary": 256,
+        "admin_origin_unrecorded": 6,
+        "unlabelled": 4,
+        "swadesh_list": 3,
+        "native_speaker_audit_session": 2,
+        "community_contributor": 1,
+        "native_speaker_confirmed": 1,
+        "wiktionary": 1,
+        "ai_prompt_enrichment": 0
+      },
+      "category_distribution": {
+        "essential": 158,
+        "places": 54,
+        "verbs": 52,
+        "adjectives": 30,
+        "numbers": 28,
+        "actions": 24,
+        "food": 21,
+        "time": 19,
+        "emotions": 18,
+        "family": 17,
+        "noun": 17,
+        "animals": 16,
+        "verb_action": 12,
+        "body": 11,
+        "professions": 10,
+        "greetings": 8,
+        "health": 8,
+        "technology": 8,
+        "clothing": 7,
+        "colors": 7,
+        "commerce": 7,
+        "nature": 7,
+        "object": 7,
+        "home": 6,
+        "education": 5,
+        "grammar": 5,
+        "particle": 5,
+        "questions": 5,
+        "adverb": 4,
+        "phrase": 4,
+        "transport": 3,
+        "verb_motion": 2,
+        "verb_state": 2,
+        "pronoun": 1,
+        "tool": 1,
+        "weather": 1
+      },
+      "difficulty_distribution": {
+        "beginner": 338,
+        "intermediate": 219,
+        "advanced": 95,
+        "expert": 1
+      }
+    },
+    "training_pairs": {
+      "total_pairs": 360,
+      "pair_directions": {
+        "sango_to_french": 360
+      },
+      "note": "Extracted from a Sango grammar and dictionary reference. Includes vocabulary entries, sentence fragments, and contextual translations. Quality varies; some entries are book fragments rather than clean parallel sentences."
+    }
+  },
+  "files": [
+    {
+      "filename": "vocabulary.jsonl",
+      "description": "Structured trilingual vocabulary entries (Sango-French-English), deduplicated",
+      "format": "JSONL",
+      "entries": 607
+    },
+    {
+      "filename": "grammar_rules.json",
+      "description": "Structured Sango grammar rules compiled from AI error analysis (13 rules, v1.1.0)",
+      "format": "JSON",
+      "entries": 13
+    },
+    {
+      "filename": "training_pairs.jsonl",
+      "description": "Sango-French translation pairs for fine-tuning",
+      "format": "JSONL",
+      "entries": 360
+    }
+  ],
+  "citation": {
+    "title": "Sango Vocabulary Dataset",
+    "authors": [
+      "MEYNG"
+    ],
+    "year": 2026,
+    "url": "https://huggingface.co/datasets/meyng/sango-vocabulary"
+  }
+}