Palestine Technical University Research Journal


The coronavirus pandemic has led to the spread of tremendous fake news and misleading information through tweets. Hence, an interesting task of classifying tweets into informative and uninformative has motivated researchers to employ machine learning techniques. The state-of-the-art studies showed high dependency on transformers architecture. However, the transformers architecture suffers from the catastrophic forgetting problem where important contextual information is being forgotten by the gradients. Therefore, this paper proposes a document embedding using Recurrent Neural Network. Lastly, three classifiers of LR, SVM and MLP have been used to classify documents into Informative and Uninformative. Using the benchmark dataset of WNUT-2020 at Task 2, LR classifier obtained the highest f-measure of 0.91. This result demonstrates the efficacy of RNN to generate sophisticated document embedding.