Typological Challenges in Treebank Construction: Morphosyntactic Mapping of the Kubeo Language in the UD Standard
Abstract
The inclusion of Amazonian indigenous languages in the field of Natural Language Processing (NLP) is hindered by the lack of treebanks and structured resources. This paper describes the initial phase of morphosyntactic modeling of the Kubeo language (Eastern Tukanoan family) under the Universal Dependencies (UD) guidelines. Kubeo exhibits high typological complexity, characterized by agglutinative processes, pervasive nominal classifiers, and pragmatic markers that challenge the static grammatical categories of the UD standard. The scope of this stage focuses on the extraction and tagging of native sentences, with a methodological emphasis on measuring the structural distance between the explicit nominal subject and the core verb. Based on data from the Pamiene Mama Buei-Tukubo primer, the study proposes token mapping solutions, utilizing the morphological features layer FEATS to preserve the language's morphosyntactic richness without excessively fragmenting the dependency trees. This pioneering corpus aims to establish a reproducible annotation protocol for future Kubeo computational tools.