Developing a Sentiment Analysis Model for Audio Datasets

In the abstract sense, sentiment analysis is an evolution arising in artificial intelligence which facilitates understanding human emotions and opinions. The motivation of this research is to investigate modelling a robust sentiment analysis on an audio dataset that has now become vital to analyze the content of the audio content such as voice recordings and podcasts. The model aims to extract, analyze and classify sentiments using high accuracy by utilizing advanced machine learning on both textual features extracted by natural language processing (NLP) and on acoustic features by audio processing algorithms. Using these methodologies, we form a multi modal approach that increases the reliability of sentiment prediction. This paper also investigates how this model performs on real data we use in benchmark datasets as well as real-life domains of interest, such as customer feedback analysis, and the monitoring of mental health and social media sentiment respectively. The suggested system grants novel avenues for assistance of understanding of a latent set of emotions in the audio content for development of new Human–Computer Interaction and automated systems applications.