Abstract: This article reports on a corpus-based study of the pronoun “se” in Brazilian Portuguese aimed at developing a proposal for annotating the different uses of the pronoun in corpora to be leveraged in Natural Language Processing (NLP). We analyzed 732 occurrences of “se” retrieved from Porttinari-base, a treebank annotated following the Universal Dependency guidelines. The results allowed us to develop a proposal for syntactic and morphosyntactic annotation, which is expected to make human annotation simpler and more consistent as well as aid supplementary annotation, as in thematic role annotation, and enhance the performance of NLP tasks, as in automatic extraction of information.