CoolFace
Datasetpublic

TigreGotico/barranquenho-ipa-dict-synthetic

Barranquenho IPA Pronunciation Dictionary The first and only IPA pronunciation dictionary of Barranquenho — the Ibero-Romance contact variety spoken in Barrancos (Baixo Alentejo, Portugal), a mixed system born of centuries of Portuguese–Spanish (Extremaduran / Andalusian) contact on the raia. Every headword is written in the Convenção Ortográfica do Barranquenho (2025) orthography and paired with a broad-phonemic IPA transcription plus Portuguese and Spanish glosses. This… See the full description on the dataset page: https://huggingface.co/datasets/TigreGotico/barranquenho-ipa-dict-synthetic.

sourceHugging Facecc-by-4.0updated 2mo agoView on Hugging Face
0likes398downloads
Dataset Card

Barranquenho IPA Pronunciation Dictionary

The first and only IPA pronunciation dictionary of Barranquenho — the Ibero-Romance contact variety spoken in Barrancos (Baixo Alentejo, Portugal), a mixed system born of centuries of Portuguese–Spanish (Extremaduran / Andalusian) contact on the raia. Every headword is written in the Convenção Ortográfica do Barranquenho (2025) orthography and paired with a broad-phonemic IPA transcription plus Portuguese and Spanish glosses.

This release supersedes the earlier LLM-synthetic version: the transcriptions are now produced by an auditable rule engine and reconciled against the scholarly and normative literature, and the lemma list is grounded in the published Dicionário de Barranquenho rather than model invention.

Columns

columnmeaning
barranquenho_orthographyheadword in the Convenção 2025 orthography
ipabroad-phonemic IPA with lexical stress ˈ at the stressed-syllable onset
part_of_speechgrammatical category (n.m., n.f., v., adj., …)
portuguese_equivalentPortuguese gloss
spanish_equivalentSpanish gloss
ipa_variantsdocumented attested alternations for this word (see below) — variation is recorded, not flattened
provenance_tierevidence tier for the IPA (see below)
sourcecitation / folio pin for the transcription
phonological_notesper-row phonological commentary

Provenance tiers

tiernmeaning
navas-attested25IPA checked against a printed transcription in Navas Sánchez-Élez (2011), folio-pinned to the printed page (and speaker tag where the monograph gives one, e.g. 3.7M.109A). Ten of these are the blind-verified orthography2ipa primary-source rows.
dicionario-headword1 483headword taken from the Dicionário de Barranquenho (2025); IPA rule-derived and convention-conformant, not individually hand-checked.
engine-verified-convention347headword from the prior dataset or the sentence gold; IPA rule-derived from the Convenção.

navas-attested is the only tier that asserts a human-verified transcription. The other two assert convention-conformance, not per-word human verification — stated honestly so downstream users can weight rows accordingly.

How the IPA is produced

Transcriptions are produced by pure [`orthography2ipa`](https://github.com/TigreGotico/orthography2ipa) — a single call to orthography2ipa.transcribe(word, "ext-PT-x-barrancos"), the declarative spec that encodes the phonology of the three coordinated 2025 normative sources. There is no auxiliary transcription layer. Earlier releases post-processed the engine output with a thin diphthong-and-stress patch; as of orthography2ipa #537 oral falling diphthongs glide (ei→[ej], oi→[oj], ou→[o], eu→[ew], …) and a word-final diphthong nucleus attracts stress (the regular Portuguese aguda rule: depois [dɨˈpoj], encontrou [ẽkõˈtro], currais [kuˈraj], fugiu [fuˈʒiw]) inside the spec itself, so that patch is now redundant and has been retired. The #537 review found that several of the retired layer's isolated-word targets in fact disagreed with the blind-judge-verified ext-PT-x-barrancos sentence gold (e.g. mesmo ˈmehmu→ˈmɛhmu, porco ˈpoɾku→ˈpɔɾku, dois ˈdojh→ˈdoj, reunião rɨˈwnjɐ̃w̃→reˈwniɐ̃w̃); this rebuild takes the pure engine form, which matches the gold, everywhere they differed.

Refresh impact: re-transcribing the 1 830 rule-derived rows (engine-verified-convention + dicionario-headword) with pure current o2i changed the IPA of 53 of them versus the prior build. Every changed row is an acute-accented falling-diphthong headword: o2i #541 now maps the Dicionário's acute offglide spellings ⟨-áu -éu -óu -íu -ái⟩ to their gliding nuclei (áu→[aw], éu→[ɛw], óu→[ɔw], íu→[iw], ái→[aj]), so abegáu [ɐbɨˈɡau]→[ɐbɨˈɡaw], berríu [bɨˈriu]→[bɨˈriw] and their ~50 siblings now carry the offglide instead of a bare vowel; the barrancos orthography round #557 on top of base #554 left the already-canonical rows unchanged. The 25 navas-attested rows keep their folio-pinned human transcriptions unchanged. On those 25 rows pure o2i independently reproduces the printed Navas form for 9 (aqui, aquilo, bonita, catorze, coluna, curso, então, pedreiro, resto) — the convergence stat holds at 9/25 across this refresh too: the acute-offglide fix touches only Dicionário headwords, none of which are Navas rows, and the remaining Navas forms preserve aspirated-coda [h] and narrow trill/tap notation the broad engine renders differently.

Modelled Barranquenho features (each cited to source in the spec):

  • —Coda-⟨s⟩/⟨z⟩ aspiration → [h] (mehmu, gohtu, rapah) — the defining trait; word-initial [s], intervocalic [z].
  • —Betacism ⟨v⟩ = [b] (binhu [ˈbiɲu], avó [ɐˈbɔ]).
  • —Final ⟨-l⟩/⟨-r⟩ loss recorded as a variant (mel [ˈmɛ], comê).
  • —Nasal monophthong ⟨-em/-en⟩ → [ẽ] (quen, nem), not the EP diphthong [ɐ̃j̃]; nasal diphthongs [ɐ̃w̃], [õj] in -ão/-ãu, -õi.
  • —Vowel reduction unstressed ⟨a e o⟩ → [ɐ ɨ u]; final atone ⟨-i -u⟩.
  • —Falling diphthongs & offglides ei [ej], oi [oj], au [aw], eu [ew], ou → [o] (monophthong), qua/gua → [kwa]/[ɡwa]; a word-final diphthong nucleus is a tonic final (aguda rule).
  • —Alveolar trill initial ⟨r⟩/⟨rr⟩ = [r] (not the EP uvular).
  • —The three-way ⟨x⟩ rule and ⟨c/g⟩ softening before front vowels.

Stress follows the Convenção defaults (paroxytone; oxytone for words ending in a diphthong or in ⟨-r/-l/-z/-n⟩), overridden by any written accent and by the lexical exception aqui [ɐˈki].

Recorded variation (ipa_variants)

Barranquenho is a living, variable, contact variety; the monograph documents speaker- and register-dependent alternation. Rather than pick one form, the dataset records it:

  • —coda ⟨-s/-z⟩: [s] ~ [h] ~ ∅ (careful / aspirated / deleted; final plural ⟨-s⟩ is often unwritten and deleted).
  • —final ⟨-e⟩: [ɨ] (dataset primary, Convenção/EP) ~ [i] (southern / Navas, e.g. catorze [kaˈtoɾzi]).
  • —⟨ei⟩: [ej] ~ [e] (retention vs monophthong; speaker-variable — pedreiro [pɨˈdɾejɾu] ~ [pɨˈdɾeɾu]).
  • —⟨-eu⟩ apocope: meu/teu/seu/eu [ew] ~ [e] (Navas 2011).
  • —lenition / syncope (narrow): agora [ɐˈgoɾɐ] ~ [aˈɣɾɐ].

91 rows carry explicit variant annotations; the alternations above apply system-wide.

Scope

Single-word lemmas only. Multiword expressions, toponyms and dictionary-internal cross-reference artifacts from the source glossary are excluded. IPA is broad-phonemic (betacism written [b], no narrow lenition diacritics in the headword form — narrow readings live in ipa_variants).

Sources

  • —Navas Sánchez-Élez, María Victoria (2011). El barranqueño. Un modelo de lenguas en contacto. Madrid: Editorial Complutense / CLUL. — the scholarly reference monograph; source of the navas-attested transcriptions.
  • —Convenção Ortográfica do Barranquenho (June 2025). Câmara Municipal de Barrancos / II Congresso Barranquenho.
  • —Gramática Básica de Barranquenho (July 2025). Gonçalves / Navas / Correia. Universidade de Évora. ISBN 978-972-778-464-6.
  • —Dicionário de Barranquenho (2025). Gonçalves / Navas / Ferreira. Universidade de Évora. ISBN 978-972-778-460-8. — headword list, glosses.
  • —Clements, Amaral & Luís (2008) on Barranquenho coda-/s/ aspiration.

Lemma list mined from `TigreGotico/dicionario_barranquenho`.

Reproducibility

retranscribe.py re-derives the ipa column with pure orthography2ipa.transcribe(word, "ext-PT-x-barrancos") for every non-navas row (the navas-attested IPA, pinned in navas_gold.py, is never overwritten). The retired auxiliary layer (bx.py) is no longer part of the build.

Limitations

  • —Non-navas-attested IPA is rule-derived, not individually human-verified.
  • —Accented offglide spellings — resolved. The spec now glides the canonical Convenção circumflex/tilde nasal finals (⟨-âu -ãu⟩ → [ɐ̃w̃], ⟨-ôi -õi⟩ → [õj]), the unaccented oral diphthongs, and the Dicionário's acute-accented offglide spellings ⟨-áu -éu -óu -íu -ái⟩ (a non-canonical variant of ⟨-âu⟩), which o2i #541 maps to [aw]/[ɛw]/[ɔw]/[iw]/[aj]. The ~50 acute rows (abegáu [ɐbɨˈɡaw], berríu [bɨˈriw], …) now carry the offglide instead of a bare vowel. This gap, the acute counterpart of the #537 fix, is closed.
  • —Rising diphthongs / `-óriu` risers. Prevocalic ⟨i⟩ before a back vowel is not modelled as a rising [ju]/[jo] glide, so -iu/-io/-óriu words are read as a falling [iw] or hiatus (belatóriu [bɨlɐˈtɔɾiw]) — an explicitly documented o2i limitation, unmodelled here rather than patched by an auxiliary layer (~26 rows).
  • —Rare learned hiatus words (e.g. reunião) may mis-syllabify.

Citation

bibtex
@misc{barranquenho_ipa_dict,
  title  = {Barranquenho IPA Pronunciation Dictionary},
  author = {TigreGotico},
  year   = {2025},
  howpublished = {Hugging Face Datasets},
  note   = {IPA via pure orthography2ipa (ext-PT-x-barrancos); gold tier pinned to Navas 2011}
}