Using Complex Deep Learning Neural Architectures for Sentiment Mining and Analysis across Text Corpora

Abstract

In order to address the issue of irony and sarcasm identification for the Italian language, we examine
and contrast five deep learning neural architectures. To determine the optimal trade-off between complexity and
performance, we quickly examine the model architectures. The reported findings, which in the best scenario
achieved 93% of the F1-Score, demonstrate how well such systems handle the challenge. As a case study, we
also demonstrate how neural systems might be integrated into a cloud computing infrastructure to take advantage
of the computational benefits of doing so when dealing with large data.