We’re excited to announce today a complete revamp of our industry-leading podcast AI technology, bringing revolutionary machine learning techniques to the audio space. I’m immensely proud of the work that the Entale team has done to make this a reality, and so pleased to be able to share it here. But first, let’s rewind a minute…
Since we launched Entale, we’ve been on a mission to make podcasts more visual and more engaging, by providing contextual content that you can interact with as you listen. That approach has proven to be a success: our users love the ‘feed style’ interaction with our content and find it to be seriously useful.
After starting with a beautiful hand-crafted CMS approach to adding contextual content to podcasts — working with podcasters like Davina McCall, Kerri Godliman and Pete Wicks to create incredible original content— we were excited to introduce the first version of our Entale AI technology last year.
From the jump, the goal of Entale AI was to provide the kind of contextual information that is relevant when you’re listening to podcasts — social media profiles of hosts and guests, Netflix links to films or TV shows being discussed — but to do it at a scale not possible with manual curation.
The first prototypes of the tech, as with all good MVPs, were built using off-the-shelf technology from Google and Amazon, and were designed to prove the feasibility of our approach. The response was overwhelming — we generated on average 3 individual items of contextual content across every single podcast listened to on the platform.
In fact, we were so happy with the outcome that we felt confident enough to hire our first data scientist to build our own, proprietary, first-of-a-kind podcast AI — and that’s what we’re releasing today.
Users who update to the latest version of the Entale iOS app today will find contextual content powered by our own proprietary AI, a neural network that is trained on podcast descriptions and designed specifically to find and display the kinds of contextual content most useful to podcast listeners. We’ve adopted a sophisticated gating approach to try and provide the most number of results as accurately as possible, whether they’re people, places, organisations, films, TV shows, works of art or events — and even on the first iteration of our tech, it’s already outperforming the Google NLP equivalent tech by 8% and producing 1.6x the amount of content per episode. Any time you can build an algorithm that outperforms something from Google, we think you’re onto a good thing…
Excitingly, this is just the beginning for our technology, which we’ve called GINA — short for Ginormous Identifying Natural language AI (yes, we love a terrible acroynm). Having built this neural network and implementation from scratch, its implementation is only going to get better, as we roll out new features including contextual content pulled from transcriptions, as well as the ability to find other instances of an entity in podcasts you don’t even follow — allowing you to follow a rabbit-hole of content based on an entity across the podosphere.
To help progress this cutting edge work, we’re also delighted to announce that we have partnered with the incredible team at Machine Intelligence Garage, as part of their latest cohort of startups, to lean on their expertise as we develop this technology further. Recognised globally as one of the top accelerators for AI-focused companies, we’re incredibly proud to be benefitting from the wealth of experience they bring to the table.
All of which leaves me with one last request — please download and give the latest version of the Entale iOS app a go, have a listen to your favourite podcast, and use the feedback button in-app to let us know what you think!