Work

Towards Uniform Meaning Representation for Brazilian Portuguese: Building a First UMR-Annotated Dataset

Year: 2026

Type: conference-paper

Abstract: This paper presents the first Brazilian Portuguese dataset annotated for Uniform Meaning Representation (UMR). It contains 96 sentences from the Brazilian Portuguese portion of the Parallel Universal Dependencies treebank, parallel to existing UMR annotations in English, Czech, and Italian. The sentences were parsed with PortParser, revised in Arborator-Grew, converted from CoNLL-U into preliminary sentence-level UMR graphs, and manually revised in PENMAN format. The results show that CoNLL-U is a useful starting point, but semantic graph construction requires manual interpretation and language-specific lexical resources. The dataset expands multilingual UMR coverage and supports future Portuguese semantic annotation.

DOI: 10.5753/stil.2026.26607

Authors: Adriana S. Pagano, Magali Duran, Federica Gamba, Daniel Zeman

Institutions: Universidade Federal de Minas Gerais, Universidade Federal de São Carlos, Charles University

Topic: Natural Language Processing Techniques

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Funders: Fundação de Amparo à Pesquisa do Estado de São Paulo, Conselho Nacional de Desenvolvimento Científico e Tecnológico, Fundação de Amparo à Pesquisa do Estado de Minas Gerais, Universidade de São Paulo

Awards: G1636247663, G769467771, G8196140990, G831127782, G9000138858

OpenAlex ID: W7222118957

FWCI:

Cites: 8

Cited by:

Related to: