g2p_barranquenho is the first open Grapheme-to-Phoneme converter for Barranquenho. Barranquenho is an Ibero-Romance contact language spoken in Barrancos, Portugal, a municipality on the Spanish border where Portuguese and Extremaduran/Andalusian Spanish have coexisted for centuries.
What makes Barranquenho phonologically interesting
Barranquenho isn’t a dialect of Portuguese or Spanish. It’s a distinct linguistic system. The Barrancos Municipal Council recently published three foundational documents: a dictionary, an orthographic convention, and a basic grammar. These gave us the rules we needed. Read the announcement: “Un Enormi Passu para u Barranquenhu i para a Cultura Barranquenha!”.
We derived the rule set from that orthographic convention. Rather than hand-rolling a bespoke pass over the text, the rules live as a language spec, ext-PT-x-barrancos, in the shared orthography2ipa engine. That spec’s grapheme table, allophone rules, stress model, and cross-word sandhi describe every Barranquenho realisation. Multi-letter graphemes collapse the way the convention documents: tch → /tʃ/, ch → /ʃ/, nh → /ɲ/, lh → /ʎ/. Nasal diphthongs appear before m/n. v always maps to /b/, and h surfaces as a pronounced /h/, unlike in either parent language.
g2p_barranquenho itself is a thin caller-side wrapper around orthography2ipa.G2P, driven by that spec. It owns text normalisation (case-folding, tokenisation into the shapes the spec expects), number expansion, and a stable phonemize/transcribe surface, but not the phonological rules. Improving a rule means editing the spec upstream, so every downstream consumer shares the fix.
In practice:
“Un Enormi Passu para u Barranquenhu i para a Cultura Barranquenha” →
ˈũ eˈnɔɾmi ˈpas̺u ˈpaɾɐ ˈu bɐrɐ̃ˈkɛɲu ˈi ˈpaɾɐ ɐ kuˈltuɾɐ bɐrɐ̃ˈkɛɲɐ
The source PDFs (convention, dictionary, grammar) are committed to the repo root so the rules are auditable against their source.
What comes next
A G2P converter is the minimum prerequisite for TTS and ASR work. Without it, a model trained on text has no principled phonetic grounding. With it, the path to a Barranquenho voice model follows the same hybrid pipeline we used for Asturian and Aragonese. The blocker is speech data, not the tooling.
If you have recordings of spoken Barranquenho or access to speakers willing to contribute under an open license, get in touch. Native speaker recordings, even a few hours, would make a TTS model viable.