Anais do XVII Simpósio Brasileiro de Tecnologia da Informação e da Linguagem Humana (STIL 2026)
Year: 2026
Type: paratext
Abstract: This paper presents the first Enhanced Universal Dependencies annotation effort for a Brazilian Indigenous language, focusing on XCOMP constructions in Nheengatu, an endangered Tupian language.We describe a semiautomatic pipeline for the UD Nheengatu-CompLin treebank to annotate control and raising predicates and propagated subjects.The method combines theoretical discussions on XCOMP constructions with corpus-based analysis of Nheengatu syntax to create rule-based enhancement heuristics.The work also provides annotation guidelines and expands EUD resources for Indigenous and low-resource languages within the Universal Dependencies framework.
DOI: 10.5753/stil.2026
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W7221966021