Large select, best practice question

Hello,
I have three tables (very simplified):
- people (~6 million records)
- addresses (~7 million records)
- contacts (~10 million records)

I need to select all records from the people table and join addresses
and contacts on it. I use the result to build a Lucene index. One
document for each person (including all contacts and addresses). I am
using jdbc.

When doing it this way, i end up with many more rows than people, and
my application logic takes care of putting the correct address/
contact with each person. To do this the records have to be sorted.
(Takes a loong time) This causes a lot of very similar result rows.
Example:

John Smith has 2 phone numbers and 2 addresses

John Smith | 555-67567 | Elm street 32
John Smith | 555-83463 | Elm street 32
John Smith | 555-83463 | P.O box 55
John Smith | 555-67567 | P.O box 55

This method is quite slow and error prone.

Is there a more elegant solution to this problem?