gvic-unb/ner-brazilian-sei-v1

Sistema Eletrônico de Informações - NER Corpus

Corpus Summary

SEI NER Corpus is a Brazilian Portuguese labeled corpus for Named Entity Recognition (NER) built from public documents of the Sistema Eletrônico de Informações (SEI), the electronic procedural management system used by Brazilian federal public agencies. The corpus was curated and validated by trained annotators and covers four document types produced by the University of Brasília (UnB): acts, leaves, resolutions, and standard announcements.

This corpus was introduced in the paper "Could you tell me the process ID? Structuring Text Documents from the Brazilian Electronic Information System Using a Named Entity Recognition Approach", presented at SBSI 2025.

Corpus Structure

The dataset is organized into four splits, one per document type:

Split	Documents	Description
acts	84	Official acts (Atos Oficiais) — designations, commissions, and administrative decisions
leaves	82	Leave authorizations and cancellations (Afastamentos)
resolutions	79	Council resolutions (Resoluções) — approvals and deliberations
std_announcements	48	Standard announcements (Editais e Retificações) — calls, rectifications, and results

Each instance contains the following fields:

Field	Type	Description
tokens	list[string]	Tokenized document text
ner_tags	list[string]	IOB2 NER tags aligned with tokens

NER Tag Set

Tags follow the IOB2 scheme (B- for beginning, I- for inside, O for outside). The full set of entity types is listed below. Not all tags appear in every split — availability varies by document type.

Tag	Entity	Description
SEI	SEI Process Number	The number of the SEI process
ORG	Organization	Organization name or acronym
PER	Person	Person's full name
LOC	Location	Location
DAT	Date	Standalone date
BDT	Begin Date	Initial date of a leave
EDT	End Date	Final date of a leave
NUM	Document Number	Document number
MAT	Enrollment Number	Civil staff identifier
SUB	Subject	Subject of the publication
UNI	University-related	Departments, secretariats, divisions, programs
OBJ	Object	Object of a contract or TED
POS	Position	Full role/title of a person
ART	Regulation/Article	Document type alongside number
DOU	DOU Information	Brazilian Official Gazette reference
MET	Meeting Number	Meeting identifier (resolutions only)
ONU	Cost	Whether there are costs to the university (leaves only)
MOT	Justification	Leave justification (leaves only)
URL	Address	Online address of an announcement
TYP	Type	Rectification or result type
WRG	Wrong Information	Information being corrected
COR	Correct Information	Corrected information

Source

Documents were collected from the public SEI bulletin of the University of Brasília (Boletim de Atos Oficiais da UnB) using a web crawler. Annotation was performed by trained annotators following a detailed annotation guide developed for each document type.

Languages

Brazilian Portuguese (pt-BR).

Licensing

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Citation

If you use this corpus, please cite:

@inproceedings{marinho2025could,
  title={Could you tell me the process ID? Structuring Text Documents from the Brazilian Electronic Information System Using a Named Entity Recognition Approach},
  author={Marinho, Mayara C and Borges, Ana Clara B and Ferreira, Laryssa O and Costa, Vanessa P and Bordim, Jacir Luiz and Borges, Vinicius R P},
  booktitle={XXI Simpósio Brasileiro de Sistemas de Informação (SBSI)},
  pages={778--787},
  year={2025},
  organization={SBC},
  doi = {10.5753/sbsi.2025.246641},
  url = {https://sol.sbc.org.br/index.php/sbsi/article/view/34395}
}