Detection of False Information in Spanish Using Machine Learning Techniques

ABSTRACT

This research focuses on the detection of false claims in Spanish through the use of machine learning techniques. Most of the current work related to automated, or semi-automated, fake news detections are carried out for the English language, however, there is still a large room for improvement in other languages such as Spanish. The detection of fake news content and its dissemination (spread) in online platforms is an open and hard problem, this work is focused on the detection of false and misleading information spreading during the election campaign in the Spanish Parliament and Catalonia crisis in 2019, migration crisis, COVID-19 pandemic, and hate speech against minorities. We propose the use of a machine learning model adapted for dealing with human language understanding tasks, called BERT, which has been trained and experimentally tested. We have collected a corpus of different types of false information and claims such as articles, posts on Facebook, WhatsApp's messages, tweets, and others. The results evidence how usage of machine learning techniques can help in the identification of false statements with more than 88% accuracy, and in collecting samples of false information. The experiments, with a comparison between different machine learning methods, have also been carried out using previous datasets, providing a comparison between different approaches. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.