According to the datamade Dedupe documentation, it seems like a gazetteer needs to have clean, distinct individual-level data.
What do you do if the individual has moved, changed jobs, etc a bunch of times? Include multiple observations per individual with the blanks intelligently filled in?
If you know that a person has had multiple addresses than I would create a 'gazetteer' like this.
Address Name Person_ID
123 Main St. John Doe 1
100 High St. John Doe 1
1600 Pennsylvania Ave John Doe 1
When you are match against this, you will have then have a second resolution step where you merge by Person_ID