Improved Sentiment Analysis Approach Using Deep Learning

Abstract

The introduction presents the importance of multimodal sentiment analysis (visual, acoustic and linguistic) and the role of deep learning in solving this problem. Then it summarizes the contribution of the thesis in solving the problem by proposing a new framework to solve the problem of multimodal sentiment analysis using generative adversarial networks. The first chapter tackles a comprehensive overview of the latest updates in the field of multimodal sentiment analysis. Then it presents a sophisticated categorization of thirty-five state-of-the-art models, which have recently been proposed in video sentiment analysis field, into eight categories based on the architecture used in each model. The effectiveness and efficiency of these models are evaluated on the most two widely used datasets in the field, CMU-MOSI and CMU-MOSEI. The second chapter tackles a comprehensive overview of generative adversarial networks (GANs) and some of its successors (DCGAN and CGAN) in order to help readers to have a panoramic view of the entire field. The third chapter gives an overview of semi-supervised learning and semi-supervised GANs. It categorizes all semi-supervised GANs, which have recently been proposed, into two categories based on the architecture used in each model, manifesting the strengths and weaknesses of each architecture. The forth chapter proposes a novel framework for multimodal sentiment analysis using semi supervised GANs, which we name "MuSA-GAN". The details the architecture of the proposed framework and the corresponding training algorithm are discussed in this chapter.