I love listening to Sommar i P1. But every year I find it difficult to discover episodes I will enjoy, and I would also like to explore older programmes.
Sommarpratkompassen is an attempt to make the programme’s archive easier to explore. The project structures information about each episode from Sveriges Radio’s API, information about each host from Wikidata/Wikipedia, and semantic topics from episode transcripts using KB-Whisper and BERTopic.
The result is far from perfect, but it already offers great opportunities to discover more interesting content and patterns in the dataset. I am releasing both the code and the data openly, and I hope it can also help researchers and other innovators explore further.
