Sentiment Analysis System for Image and Text Based Social Media Data

The paper provides an explicit outlook on the development of a comprehensive sentiment analysis systems for online social media by targeting user-generated text and images on platforms such as Twitter, Facebook, and Instagram. The system is developed to provide a complete sentiment analysis of users' emotions and expressions on social media. The project is composed of three interconnected sections, a Text-based sentiment analysis model, an Image-based sentiment analysis model and a web application prototype that integrates these models to provide real time results. The text-based model includes, data preprocessing, analysis using Word Cloud, analysis using the techniques of Word2Vec and implementation of multiple machine learning and deep learning models. Similarly, image-based model includes data preprocessing, data augmentation, deep learning methods for sentiment assessment. Final module is focused on representing application with real time demonstration, consisting with the practical implementation of these models. The prototype displays a real-time sentiment analysis and visualizes data. The paper is focused on a wide range of the field where the importance of the sentiment analysis is immense. For instance, companies can analyze how products are being analyzed in the market, politicians can determine how people are thinking of their speeches, teachers can know the interest of the lecture for the students. Therefore, the research signifies the importance of the Sentiment Analysis to extract the meaning or insight from the overwhelming unstructured, dynamic and always evolving nature of the social media content.