The data dump format is described here. It's changed some since I wrote the answer that you reference, but the basic principals are still the same.
The MQLread and Search APIs can both be used to filter Freebase topics to a specific type (e.g. the /people/person type). Topics can have multiple types, but if what you want is the classic NER types, then the first thing to do is probably filter by the desired types since there are "only" about 3 million people in Freebase out of the tens of millions of topics.
Freebase contains a lot more information than just the name which would be useful in an NER task, so you might want to take a step back and think about what other properties you might want to extract as well.