yijingyidong / demo_inpe_covid

This repo contains code and corpora for the twitter robot-journalists @DaMataNews, @DaMataReporter and @CoronaReporter described at the INLG 2020 paper entitled "DaMata: A Robot-Journalist Covering the Brazilian Amazon Deforestation" and at the ENIAC 2020 paper entitled "Towards Fully Automated News Reporting in Brazilian Portuguese".

This repository contains code and corpora for the twitter robot-journalists @DaMataNews, @DaMataReporter and @CoronaReporter. This work was described in the following articles:

DaMata: A Robot-Journalist Covering the Brazilian Amazon Deforestation

INLG 2020 - 13th International Conference on Natural Language Generation
URL: https://www.aclweb.org/anthology/2020.inlg-1.15

Towards Fully Automated News Reporting in Brazilian Portuguese

ENIAC 2020 - XVII Encontro Nacional de Inteligência Artificial e Computacional
DOI: https://doi.org/10.5753/eniac.2020.12158

Requirements

Dependencies to run the code may be installed by the following command:

pip install -r requirements.txt

Database

Non-linguistic data about COVID-19 in Brazil and Deforestation in the Legal Amazon area extracted from the web and stored in a private MongoDB database, accessed by the MongoEngine ORM framework. So in order to run the code, make sure to set up a MongoDB database and update the login info on the db/model.py file. Source data for COVID-19 can be retrieved from Worldometers website and data for Amazon deforestation can be retrieved from INPE Terrabrasilis platform.

Paths

In the paths.py file, you will see the path for several important files, such as the structuring, lexicalization, references grammas as well as the lexicon and the Twitter API login info. In this file, make sure to add the proper files and keys before run the code.

Execution

Once all the dependencies are solved, the robot-journalist can be tested on the Jupyter files covid19.ipynb and deter.ipynb. To be executed on a production environmente, the publisher.py files for COVID-19 and Deforestation can be executed.