Opinion Mining from Audio Conversation using Machine Learning tools and BART Transformer

The discipline of opinion mining is expanding rapidly. It gathers subjective data from text and audio sources using Natural Language Processing and text analysis. Finding and extracting subjective data automatically from news stories, blogs, social media posts, customer evaluations, and other sources is the primary objective of opinion mining. Opinions, assessments, appraisals, sentiments, and emotions fall within this category. As consumers increase and data accumulates daily, companies rely on data collection and analytics to track sales, customer insights, and brand reputation. However, voice data, the most direct feedback, is often overlooked. This paper explains how to analyse the sentiment in audio recordings and apply the top three approaches in three different groups of opinions (positive, negative, and neutral) in order to gain a better understanding of how customers evaluate goods and services. An extensive examination of benchmarked datasets, popular feature sets, algorithms, approaches, open-source tools, difficulties, practical applications, and insights into several facets of opinion mining are provided in this research. With 99%+ accuracy for various data sets using Python's BART transformer, these results have both theoretical and practical implications, as well as demonstrating the utility of opinion mining from sound to text converted information in a holistic society.