g2p-barranquenho 0.2.1a1

g2p_barranquenho - Barranquenho Phonemizer

This repository holds a grapheme-to-phoneme (G2P) converter for the Barranquenho language. Give it written Barranquenho text and it returns IPA.

Barranquenho is an Ibero-Romance variety spoken in the municipality of Barrancos, Portugal. It shares several features with the neighbouring Extremaduran and Andalusian varieties of Spanish: coda-/s/ aspiration, betacism (v becomes [b]), the alveolar trill, and southern vowel reduction.

Install

pip install -e .

This pulls in orthography2ipa automatically.

Architecture

The phonology lives in the shared orthography2ipa language spec ext-PT-x-barrancos. That spec holds the grapheme table, the allophone rules, the stress model, and the cross-word sandhi rules. It is the single source of truth for how Barranquenho is realised. This package is a thin wrapper around orthography2ipa.G2P, driven by that spec:

• transcribe(text) phonemises a whole utterance, so the spec's cross-word sandhi applies (coda-/s/ aspiration and deletion, article and conjunction destressing, vowel elision at word boundaries).
• phonemize(word) phonemises one word and returns its phones as a list.
• The package handles only caller-side text handling (case and Unicode normalisation) and this stable surface. The linguistic rules belong to the spec, so a rule improvement is an upstream edit that every consumer shares.

Lexical stress is marked with ˈ before the stressed syllable. Nasal vowels carry the combining tilde U+0303 (ɐ̃). A coda m or n is absorbed into the preceding nasal vowel rather than surfacing as its own segment.

Usage

phonemize takes one word and returns a list of IPA phones. Join the list for a compact string.

from g2p_barranquenho import phonemize

phonemize("boca")      # ['ˈb', 'ɔ', 'k', 'ɐ']
phonemize("manhán")    # ['m', 'ɐ', 'ˈɲ', 'ɐ̃']
phonemize("ambu")      # ['ˈɐ̃', 'b', 'u']

"".join(phonemize("cahtelu"))   # 'kɐˈhtɛlu'

transcribe takes a whole utterance and applies cross-word sandhi.

from g2p_barranquenho import transcribe

transcribe("O tempu não ehtá nada bom agora.")
# 'o ˈtẽpu ˈnɐ̃w̃ eˈhta ˈnadɐ ˈbõ ɐˈɡɔɾɐ'

transcribe("Comprámos pão e vinho na feira de Barrancos.")
# 'kõˈpɾamu ˈpɐ̃w̃ i ˈbiɲu nɐ ˈfejɾɐ dɨ bɐˈrɐ̃ku'

The BarranquenhoG2PPlugin class in g2p_barranquenho.plugin exposes the same engine behind the string-returning transcribe and transcribe_word methods other components in the toolchain expect.

Numbers

Digits carry no orthography the lattice can read, so the normalizer stage spells numeric tokens out into Barranquenho words before transcription. This runs by default in transcribe. Pass expand_numbers=False to leave digits as they are.

from g2p_barranquenho import transcribe
from g2p_barranquenho.number_utils import normalize_numbers, BarranquenhoNumberParser

transcribe("tenho 3 gatu")                       # 3 -> 'treh', then transcribed
transcribe("tenho 3 gatu", expand_numbers=False)  # digit left to the spec

# spell numbers to text without phonemising
normalize_numbers("tenho 20 anu")                # 'tenho binti anu'
normalize_numbers("la casa 1ª")                  # 'la casa primeira' (º masc / ª fem)

# the verbaliser directly
p = BarranquenhoNumberParser()
p.cardinal(256)              # 'duzentuh e cinquenta e seih'
p.cardinal(2, "feminine")    # 'duah'
p.ordinal(1, "feminine")     # 'primeira'
p.pronounce_token("3,5")     # 'treh birgula cincu'

Numeral groups join with the copulative e ("and"). Navas Sánchez-Élez (2011) documents no numeral paradigm, so most forms come from the Convenção Ortográfica do Barranquenho (2025) spelling rules: final -o becomes -u, final -e becomes -i, coda -s is written -h, and v becomes b. The lexemes Navas does attest (2 douh, 3 treh, 6 seih, 7 seti, 10 deh, 14 catorzi, 20 binti, 100 cien, 1000 mil, 1st primeiru) live in number_utils.ATTESTED. Everything else lives in number_utils.DERIVED, so a reader can tell citation from reconstruction.

Documentation

• docs/quickstart.md - install and the first call
• docs/api.md - the functions, their contracts, and the plugin
• docs/advanced.md - stress, sandhi, diacritics, and recipes
• examples/ - runnable scripts

Related projects

• OpenVoiceOS/orthography2ipa - the upstream G2P engine and language-spec system this package drives

Sources

The ext-PT-x-barrancos spec derives from the coordinated normative suite produced under the Programa de Preservação e Valorização da Língua e Cultura Barranquenhas, plus the phonetic descriptions of Navas Sánchez-Élez:

• Convenção Ortográfica do Barranquenho (June 2025). Câmara Municipal de Barrancos / II Congresso Barranquenho working group.
• Gramática Básica de Barranquenho (July 2025). Maria Filomena Gonçalves, María Victoria Navas, Victor M. Diogo Correia. Universidade de Évora, 1ª edição. ISBN 978-972-778-464-6.
• Dicionário de Barranquenho (2025). Maria Filomena Gonçalves, María Victoria Navas, Vera Ferreira. Universidade de Évora, 1ª edição. ISBN 978-972-778-460-8.
• Navas Sánchez-Élez, M. V. (2011). El barranqueño: un modelo de lenguas en contacto. Madrid: Editorial Complutense.

License

Apache-2.0. See LICENSE.