Name Entity Recognition for Brazilian Portuguese.

The present repository is a study of Name Entity Recognition application to detect proper nouns in Portuguese.

The study was carried on CE-DOHS corpus (Corpus Eletrônico de Documentos Históricos do Sertão).

CE-DOHS was preprocessed and, later, annotated using label-studio. This study aimed to calculate the NER F-Score using BI-LSTM-CRF deep learning algorithm.

All the process can be found in Full Pipeline notebook.

The F-Score obtained was 0.97

How to use: Notebook version

1. conda create -n NameAnonPT
2. conda activate NameAnonPT
3. pip install -r requirements.txt
4. jupyter lab

How to use: Airflow Version

OBS: this version requires Docker and Docker-Compose.

1. Go to Airflow folder
2. docker-compose up
3. Access localhost:/8080
4. run cartasAnonPT DAG

Below some screenshots: