Multilingual Sentiment Analysis through Integrated Multimodal Deep Learning Techniques

Abstract

Sentiment Analysis (SA) is one of the prominent area of research nowadays. Researchers contribute alot to extract sentiments from text, audio and video. This paper focuses on all three modalities such as text, audio, and video to enhance SA. It contributes various methodologies and techniques for SA across different languages and areas. In this paper, Machine Learning (ML), Deep Learning (DL) and Model Fusion (MF) techniques for extraction of features were discussed. This paper also contributes the literature from different researchers about clustering application and extraction of sentiments using text, audio and video. The paper the role of labeled and unlabeled data for development of multimodal system. Challenges to detect sentiments and emotions from literature, rumors and fake news also discussed in detail. Paper contributes the methods and techniques produced high accuracy and precision of SA using various languages and domains. This paper helps the researchers to analyze methods and techniques used to extract sentiments from different languages and domains using text, audio and video namely tri-model sentiment architecture.