Rate and Review – A Replicable Method For Podcast Research
Along with my BCMCR colleague, Dr Simon Barber, I’m currently writing a journal article around the use of podcasts in contemporary music reception practice.
The research behind the article has involved the collection and analysis of data related to podcasts in the Apple Podcasts charts for music. During the course of undertaking this work, I have developed a replicable workflow for gathering, analysing and visualing that data.
For the purposes of the article we are writing, we gathered data on ~9,000 podcast episodes and ~16,000 reviews, and then ran the data through a number of unsupervised machine learning algorithms, including Topic Modelling and Sentiment Analysis. The results are displayed in this interactive document which has enabled us to explore the contents of the reviews and the overall results of computational analyses.
The article is currently in peer review and we hope to have positive news on that soon. In the article, we make the case that the replicable workflow developed as part of this research has potential uses for other researchers interested in podcasts. Towards that, I have created a series of tutorials that provide walkthroughs and all code required to replicate to workflow we used. These are available over on my personal website:
- Part 1 – Describes the collection of data.
- Part 2 – Walks through the process of analysing the data using Topic Modelling and Sentiment Analysis.
- Part 3 – Visualises the results of those analyses.
- Part 4 – Describes how to create an interactive document enabling exploration of results.
If you are interested in researching podcasts, please feel free to use, share and adapt any of the code or processes contained within the tutorial series.