Summary
Nheengatu is an Amazonian language with few speakers and little translation technology available. The authors gathered the largest collection of Portuguese-Nheengatu text pairs, including the Brazilian Constitution translated into Nheengatu, and built open machine translation models for both directions. They found that translating legal text into Nheengatu is difficult due to inconsistent legal terms but can improve by using a fixed glossary during training and decoding. Adding transcribed conversations to training data boosts translation quality on spoken language but not on written text. Larger general models help translating into Portuguese but not into Nheengatu for legal or speech content.
What this means in practice
- •For language technology developers: Train and evaluate Portuguese-Nheengatu translation tools using the first open corpus and models including legal and spoken domains.
- •For multilingual service providers: Improve translation quality for Nheengatu speakers in legal and conversational settings using technologies tailored to the language's challenges.$Commercial implications: Facilitates delivery of professional translation or localization products for a minority language with newly available domain-specific resources and models.
Authors
Fabrício Carraro, Rodolfo Joel Zevallos, Rodrigo Gonçalves de Souza, Caio Henrique Faustino da Silva, John E. Ortega
Abstract
Nheengatu (yrl), the Amazonian Língua Geral, has 6--10 thousand speakers and is co-official in São Gabriel da Cachoeira (Brazil), yet remains poorly served by evaluated Portuguese--Nheengatu MT resources. We present the largest parallel corpus assembled for the language (14.5k Portuguese--Nheengatu pairs), anchored by the 2023 official translation of the Brazilian Federal Constitution; the first MT evaluation set built from transcribed spontaneous speech; and, to our knowledge, the first open-weight MT models dedicated to bidirectional Portuguese--Nheengatu translation. Three findings organize the paper. First, translation into Nheengatu on held-out constitutional articles plateaus on development data regardless of training length or model size, and much of the remaining error sits in conventional legal terms that the written sources do not attest consistently; a terminology success rate over a frozen lexicon shows that glossary mechanisms at training time and lexical constraints at decoding each raise term realization, the latter at a cost in repetitive output that chrF++ barely registers. Second, a corpus expansion that adds transcribed conversation raises speech chrF++ by 24--27 points while leaving written domains unchanged, a register gap that written-only benchmarks cannot see. Third, a larger NLLB model helps translation into Portuguese but not into Nheengatu on legal or speech text. For Portuguese-to-Nheengatu translation, our final systems reach 31.7--32.8 chrF++ on article-disjoint constitutional text, 64.1--64.5 on the near-domain set (62.1 after removing items with target-side content overlap), and 45.2--47.5 on speech.