It's What You Say and How You Say It: Exploring Audio and Textual Features for Podcast Data

Neil Kumar Shah, Vivek Srivastava, Mohit Bhardwaj, Satej Kadlay, Dharmeshkumar Agrawal, Savita Bhat, Niranjan Pedanekar

Abstract: Podcasts are a relatively new media type, consisting of spoken documents or conversations with a wide range of topics, genres, and styles. It is of utmost importance to understand what makes a podcast more popular than others or which tags help to characterize a podcast. Here, we provide a comprehensive analysis of widely used traditional features from two modalities, i.e., audio and text. We explore multiple feature combinations considering podcast popularity prediction and multi-label tag assignment as proxy downstream tasks. Our analysis is focused on examining prevalent hypotheses and identifying new perspectives about the podcast content.

External IDs:dblp:conf/apsipa/ShahSBKABP23