Improving Information Accuracy

Over the past two weeks, we’ve been working on improving the accuracy of live info presented to listeners, investigated a case where a song was incorrectly being identified, did some studio-side maintenance, and continued to work on new podcast hosting. For more details, read on.

The accuracy of song information is one of the most important things the station provides beyond the radio stream itself. Since the radio’s inception, we’ve grown through several different radio platforms which have all used slightly different ways of providing data about songs that play on the station. At its core, we use an API provided either by the platform or by other cues which we then send over to our web server to present on our website, app, and our own public API which is used by some third-party applications. It’s a dance which is made more complex by variances in metadata, character encoding, file formats, and data transfer.

Recently, we remade part of our live song identification system which, over time, had become quite complicated to handle different cases, and we noticed it would misidentify songs in specific cases. We’ve since cleared misidentified songs from the past few months which also resulted in a small handful of ratings and favorites being cleared.

We simplified the system in a way where songs won’t be misidentified in this way again, but now we have the inverse problem where some songs are not identified at all, causing song ratings and favorites to become unavailable for those songs. We’re continuing to work through adding internal methods by which songs are identified while minimizing the risk of misidentification to reduce the number of non-identified songs.

Finally, we upgraded a couple systems studio-side and are reimplementing some supporting systems for the station such as music rotation. The station is not impacted by these systems being offline for a short period of time.

We continue to test our new podcast hosting solution before making episodes available once again, so stick around for more news soon, and thanks for listening!

[Knowledge #200]