Identification of COVID-19 Related Fake News via Neural Stacking

Abstract

Identification of Fake News plays a prominent role in the ongoing pandemic, impacting multiple aspects of day-to-day life. In this work we present a solution to the shared task titled COVID19 Fake News Detection in English, scoring the 50th place amongst 168 submissions. The solution was within 1.5% of the best performing solution. The proposed solution employs a heterogeneous representation ensemble, adapted for the classification task via an additional neural classification head comprised of multiple hidden layers. The paper consists of detailed ablation studies further displaying the proposed method's behavior and possible implications. The solution is freely available. https://gitlab.com/boshko.koloski/covid19-fake-news

Implementation of the BERT-derived architectures to tackle disinformation challenges

“…Although no results were found in the literature for the ISOT dataset used in the article, for similar datasets and similar challenges related to the detecting fake news, the authors in articles [32] and [1] obtained results of a similar range. The COVID-19 dataset, related to the competition under CONSTRAINT 2021 [25] [26], was the subject of research in [13] [22] [16] primarily in the range of f1-score metrics. The best results obtained in this study, in relation to the f1-score, exceed all those reported so far.…”

“…Various techniques are used to improve the quality of the metrics. The authors of the article [16], in order to maximize the f1score metric, adopted, among others, a strategy of constructing handcrafted features that captured the statistical distribution of words and characters [16]. In turn, in the work [22] to maximize the metrics, pseudo-label algorithm and text-transformers architecture, consisting of five different Transformers models (BERT, Emie, RoBERTa, XLnet, Electra), were applied [22].…”

“…The focus was on two columns, article 'title' and 'text', and the models were created for the data contained therein. The COVID-19 dataset consists of three files that correspond to three sets for training, for validation [22] and for testing [16]. The sets contain classified data divided into two categories.…”

“…2) is proposed, which is based on Transformers; the analysis and training of the neural network were extended with a dataset containing fake and real news from the COVID-19 area; a detailed description of the encoders included in BERT and RoBERTa has been added; additional charts related to the pre-experiment procedure, which were conducted to maximize metrics (Figs. 13,14,15,16,17), have been prepared.…”

Detection of Fake News Text Classification on COVID-19 Using Deep Learning Approaches

“…After classification experiments, LR achieved 75.65% accuracy, embedding with dense layer achieved 86.93% accuracy, embedding with LSTM layer achieved 86.9% accuracy, and bi-LSTM model achieved 72.31% accuracy. [52] states in their research that fake news have important role in everyone's life in these days. An individual's life can totally change due to these fake COVID-19 news.…”