This week, we spent much of our time putting together a new system which aims to exceed the current scheduler’s capabilities in terms of automated playlist construction. For more details, read on!
When we began co-locating our own server hardware over two years ago, we decided to use a specialized piece of software which has driven the station’s playlist since its deployment. It has allowed us to craft customized queries and build our playlist how we wanted, with some caveats. The system has largely been isolated from any signals that come from listeners such as ratings, favorites, changes in listener count, or who’s actively listening.
The new scheduler we’re building to replace the current one is based on a series of scripts and signals which will grow to include those mentioned above. Right now, the focus is on matching the current scheduler’s capabilities, but we are looking forward to building in an adaptive system which curates the live playlist based on who is currently listening and their song preferences. The goal is to emphasize songs people are likely to enjoy, repress songs the current listener group is less likely to enjoy, and create the best station we can for any given group. We will still bring in new songs, songs with fewer ratings, and songs which aim to expand what listeners might normally listen to on their own.
We’ve been talking about an adaptive system for a number of years at this point, but with this next generation scheduler, we’ll finally make that a reality. With that said, building this has not been easy so far, and we’ll likely be iterating and improving it over time just as we had when the current scheduler was first implemented.
That’s all for this week, tune in this weekend for our next Live show, and thanks for listening!
[Knowledge #155]