Podcast Episode Details
Back to Podcast Episodes
Apple Podcasts is using machine learning to tag episodes and 7 other stories
Episode 37
This Week: Apple Podcasts is using machine learning to tag episodes, IAB to require annual recertification, Edison Research shares Share of Ear Q3 statistic, and Headliner announces automated YouTube integration. Bumper Discovers Apple Podcasts Assigns Topics Automatically. Manuela: This Tuesday Bumper co-founder Dan Misener posted a new discovery that explains how Apple Podcasts can recommend individual podcast episodes based on topics discussed. The example Misener uses is an episode of Today, Explained which, if one opens the web page source on Apple Podcasts, is tagged with twenty topics ranging from broader concepts like ‘world politics’ to individual names of celebrities and politicians mentioned in the episode. The catch? Those keywords and topics do not appear anywhere in the episode description or RSS feed. The only way to assign them to that particular episode is a transcript. From the article: “Here’s my best guess: Apple is using machine-generated transcriptions, then applying natural language processing techniques like topic modeling to generate lists of relevant topics on an episode-by-episode basis.” According to Misener’s reporting, the current top 250 podcasts on Apple Podcasts consists of 70,094 episodes. Approximately 63.5% of those episodes are currently tagged with topics generated by Apple. “Here’s my best guess: Apple is using machine-generated transcriptions, then applying natural language processing techniques like topic modeling to generate lists of relevant topics on an episode-by-episode basis.”
The topics are also ranked with a per-episode relevance score and appear to be integrated into the Apple Podcasts search function. Misener tested this by searching the phrase ‘war in Donbass,’ which he had seen as a tag on Today, Explained. Apple Podcasts returned an episode of The Inquiry that discusses the issue at length, but also does not specifically include those keywords in the title or description. This suggests the assigned topics influenced its search ranking. As Misener says in his breakdown of what this means for podcasters, the implications of this automated topic system are numerous and all signal better relevancy in podcast discovery. SEO now goes far beyond what they chose to include in RSS feeds. In response to the Bumper article, https://podnews.net/article/apple-topics. Visitors can type in the name of a podcast and choose one of the 20 most recent uploads to see what topics Apple has assigned. That is, if it has been assigned any at all. Unfortunately the Sounds Profitable feed has not been fed through their machine learning, so we don’t know what Apple thinks last week’s episode of The Download is about. IAB to require annual recertification. Shreya: Last Thursday Podnews reported the Interactive Advertising Bureau is now requiring annual re-certification from podcasting companies. The information came from a note Podnews editor James Cridland discovered on the IAB website.
https://iabtechlab.com/standards/podcast-measurement-guidelines/“As podcast listenership increases and the technology to support that listenership improves, the podcast technical measurement capabilities are continuing to evolve at a rapid pace. As such, and to be aligned with other industry auditing programs, IAB Tech Lab is updating its compliance program to require annual recertification. “
The post then cites the fact several complaint companies were certified on the 2.0 version of the standards but have not re-certified under 2.1. As of this podcast https://iabtechlab.com/compliance-programs/compliant-companies/#\ For context: version 1 of the Measurement Technical Guidelines was released in September of 2016. Version 2.
Published on 2 years, 5 months ago