Predição de popularidade de podcasts através de características textuais

Abstract

With the tremendous growth of Podcasts and the professionalization of its creators,to the point that news networks call this as Podcast’s ”golden age”, new tools have emerged to assist its content producers in building and maintaining of their channels.In this context, finding features inside episodes that provide broader reach to the target audience is of great value to both creators and listeners, allowing channels to stay active longer and offer better content quality.Thus, this paper proposes a study of popularity analysis of brazilian’s podcasts using a podcast audience analysis tool in one of the most used channel and episode aggregators in the world, iTunes. By using Web Scraping tools to collect available and necessary information, also tools for transcriptions of the audios’s episodes in orderto obtain what has been said, and calculating metrics to measure the accuracy of the generated model, therefore making an analysis of which information is relevant or not o predicting a channel’s popularity.Results displayed were favorable in the correlation between the categories analyzed individually and the its text, whereas in an analysis in which categories are not discriminated there is a low relationship between text and popularity, demonstrating that the category of a given channel plays an important role in analyzing its popularity.