Yeah, I need to find a better way to handle the (very common!) case where a podcast doesn't yet have enough highly-ranked similar shows. Similiarity is based o

Yeah, I need to find a better way to handle the (very common!) case where a podcast doesn't yet have enough highly-ranked similar shows. Similiarity is based on a mix of some basic nlp (tf-idf similarity between show and episode notes) and shows that users have starred together or added to a multi-search. I initially tried lowering the baseline score for what constitutes a 'similar' show, but for shows where I don't have any non-nlp features it often produced garbage.