recogna-nlp/fakerecogna2-extrativa

FakeRecogna 2.0 Extractive

FakeRecogna 2.0 presents the extension for the FakeRecogna dataset in the context of fake news detection. FakeRecogna includes real and fake news texts collected from online media and ten fact-checking sources in Brazil. An important aspect is the lack of relation between the real and fake news samples, i.e., they are not mutually related to each other to avoid intrinsic bias in the data.

The Dataset

The fake news collection was performed on licensed and verified Brazilian news websites with enrollment in the Duke Reporters´ Lab Center. The system was designed as a source to fight against fake news spreading worldwide. For real news, we selected well-known media platforms in Brazil. Since real texts are much larger than most of the produced fake content, the genuine news was preprocessed with text summarization. At this stage, there is no further processing of stop words or lemmatization of the text. After trimming and standardizing the real news, we produced textual representations based on Bag of Words (BoW), Term Frequency – Inverse Document Frequency (TF-IDF), FastText, PTT5, and BERTimbau to form the input feature vectors for the ML models. Figure illustrates the steps of the proposed method.

Fake news sources were selected from nine fact-checking agencies in Brazil. This process provides a broad range of categories and many fake news samples to promote data diversity. Table 1 presents the existing Brazilian fact-checking initiatives and the number of fake news samples collected from each news source. When the search process was concluded, we ended up with 26,569 fake news samples, which, in turn, were further processed to detect and remove possible duplicate samples, thus leading to a final set of 26,400 fake news articles.

Fact-Check Agency	Web address	# News
AFP Checamos	https://checamos.afp.com/afp-brasil	1,587
Agência Lupa	https://piaui.folha.uol.com.br/lupa	3,147
Aos Fatos	https://aosfatos.org	2,720
Boatos.org	https://boatos.org	8,654
Estadão Verifica	https://politica.estadao.com.br/blogs/estadao-verifica	1,405
E-farsas	https://www.e-farsas.com	3,330
Fato ou Fake	https://oglobo.globo.com/fato-ou-fake	2,270
Projeto Comprova	https://checamos.afp.com/afp-brasil	887
UOL Confere	https://noticias.uol.com.br/confere	2,579
Total	-------------------------------------	26, 569

More informations

The FakeRecogna 2 dataset is a single XLSX file that contains 8 columns for the metadata, and each row represents a sample (real or fake news), as described in Table 2.

Columns	Description
Title	Title of article
Sub-title (if available)	Brief description of news
News	Information about the article
Category	News grouped according to your information
Author	Publication author
Date	Publication date
URL	Article web address
Label	0 for real news and 1 for fake news

FakeRecogna v2 - Abstrative

The abstrative summarization version of FakeRecogna 2 can be found here.

Citation

@inproceedings{garcia-etal-2024-text,
title = "Text Summarization and Temporal Learning Models Applied to {P}ortuguese Fake News Detection in a Novel {B}razilian Corpus Dataset",
author = "Garcia, Gabriel Lino  and
  Paiola, Pedro Henrique  and
  Jodas, Danilo Samuel  and
  Sugi, Luis Afonso  and
  Papa, Jo{\~a}o Paulo",
editor = "Gamallo, Pablo  and
  Claro, Daniela  and
  Teixeira, Ant{\'o}nio  and
  Real, Livy  and
  Garcia, Marcos  and
  Oliveira, Hugo Gon{\c{c}}alo  and
  Amaro, Raquel",
booktitle = "Proceedings of the 16th International Conference on Computational Processing of Portuguese - Vol. 1",
month = mar,
year = "2024",
address = "Santiago de Compostela, Galicia/Spain",
publisher = "Association for Computational Lingustics",
url = "https://aclanthology.org/2024.propor-1.9/",
pages = "86--96"
}