Recommendations: with a graph database for the ratings

Because of the limitations of a relational database, the recommendations are currently based on two or three films we've liked.

With a graph database, it would be possible to connect each user to each film they've liked. Then, it would be a simple matter to search the users who have the most connections in common with us.
That way, it would be possible to find our "film taste twins" and get better recommendations and even rating predictions.

See http://pizzaconnections.net/2015/03/27/148/ for an example.