Developing a Hybrid Approach for Enhanced Sentiment Analysis Integrating Textual and Audio Data Streams

Abstract: We are living in the era where social media plays a vital role. Online social networking sites like Facebook, YouTube, and Twitter have gained popularity as the number of social media technologies has expanded because they enable people to discuss and express their ideas about numerous life events. The bulk of people spend most of their time on social media sites every day. Using a dataset of 27481 records from Kaggle, we trained our deep learning model. We predict the sentiment into 3 classes with positive, negative or neutral polarity for the opinions expressed in the form of either text or audio. Additionally, our proposed technique has various practical applications and improves the accuracy of sentiment prediction.

Keywords: Sentiment Analysis, Neural Network, Natural Language Toolkit (NLTK), Twitter sentiment analysis, Natural Language Processing (NLP), Text based Sentiment Analysis