How to exploit big data from social networks: a subjective well–being indicator via Twitter