INTEGRATING AI AND CLOUD COMPUTING FOR EFFICIENT AUDIO ANALYSIS

Abstract

Abstract. This paper propose a pipeline for sentiment analysis of audio inputs by using AI and techniques for speech recognition. A key aspect of our approach involves determining audio chunks based on silence detection and decibel height, allowing for more effective segmentation and analysis of the input data. These techniques will be instrumental in categorizing audio materials, enabling more organized and accessible content management. In the scope of this paper, we will compare and utilize several Python packages for audio detection, including Librosa, PyDub, and SoundFile, which offer various functionalities for audio analysis and manipulation. Additionally, we discuss the high computational demands associated with training and deploying of models, highlighting the potential cost barriers for many organizations. To address these challenges, we propose the use of cloud-based solutions, specifically AWS Spot Instances, Azure Spot VMs, and Google Cloud Preemptible VMs, which offer substantial cost savings for processing audio data. By leveraging these cloud resources, organizations can significantly reduce expenses while maintaining high performance in AI audio processing tasks.