Things we’ve noticed from looking at real data that we could handle in a
smarter way:
- removing numbers (there might be a tom.smith2@dept.gov.uk if tom.smith
is already taken)
- removing middle initials (again, these tend to be used for
disambiguation and aren’t included when we ask people for their names)
- ignoring email addresses which only have someone’s initial, not their
first name (because we can’t make a decent guess in this case)
660fadbce7