MUBI's curatorial approach stands in sharp contrast to major streaming services
like Netflix, Hulu, and Amazon, which have amassed large libraries and deliver
personalized recommendations based on algorithms. MUBI's human-curated selection
takes a lot of that choice away, and Cakarel thinks viewers are better off
for it.
"Think about your own Netflix experience and how frustrating it is — how
long it takes you to find a film that you want to watch," says Cakarel.
"It doesn't work. It categorically doesn't work."
While I get what they're saying, I still wonder if it wouldn't be interesting to try out some personal recommendation algorithms/systems on their user ratings dataset. (After reviewing their T&C, I've actually crawled through their data (more like, their backend exposes neat API endpoints that are not part of any formal API (officially they don't have one)..) some time ago, and am still curious to just try out some stuff from scikit-learn.
Maybe someone has thought/done the same? When contacted, the MUBI team (IIRC) basically said they're not interested in that as of now.
Maybe someone has thought/done the same? When contacted, the MUBI team (IIRC) basically said they're not interested in that as of now.