Now well engrossed in the old family tree I was thinking about all the useful advice and suggested information sources I had received (often from friends here).
So far my tree goes back to pre 1850 for all four major lines (the grandparents) and I’m 99.9% happy all are correct (and that based on just two marriage certs.). But when going sideways (into parts of the tree I didn’t previously know existed) or prior to say 1837 the confidence factor goes down.
Sometimes of course you get lucky and get circular references. My great grandma has a daughter. The daughter goes through two marriages. The mother is already linked to two other siblings. Many years later the great grandma with one sibling in tow moves in with her daughter. Without boring you with all the details I have two distinct trails which go their separate ways and then reunite at a later date. The chance of this not being great grandma is now very slim.
Well I was thinking, being that way inclined, of creating a formula to calculate confidence based on quality of data, number of data items, number of sources, quality of reference, independence of source (a lie maintained becomes fact), etc.
What really got me thinking along these lines, and I don’t want or intend to be critical here, was some of the crazy connections ancestry offered for members of my tree. Now generally they are useful but at this simple level I feel they could be a bit smarter. (I’m in IT development by the way so I know how difficult developing clever software using what is known as ‘fuzzy logic’ is. And IMHO Ancestry as an example of a website is very slick).
But before I get started I wondered if anyone had seen any research into this aspect of our hobby? Any references might save me reinventing the wheel.
Cheers Nigel
p.s. I know the detection is most of the fun and I don’t aim to look to automate that. I just am interested in generating probability of accuracy. For example an algorithm might examine a tree documented with sources and identify weaknesses that warrant further verification.
Cheers Nigel
So far my tree goes back to pre 1850 for all four major lines (the grandparents) and I’m 99.9% happy all are correct (and that based on just two marriage certs.). But when going sideways (into parts of the tree I didn’t previously know existed) or prior to say 1837 the confidence factor goes down.
Sometimes of course you get lucky and get circular references. My great grandma has a daughter. The daughter goes through two marriages. The mother is already linked to two other siblings. Many years later the great grandma with one sibling in tow moves in with her daughter. Without boring you with all the details I have two distinct trails which go their separate ways and then reunite at a later date. The chance of this not being great grandma is now very slim.
Well I was thinking, being that way inclined, of creating a formula to calculate confidence based on quality of data, number of data items, number of sources, quality of reference, independence of source (a lie maintained becomes fact), etc.
What really got me thinking along these lines, and I don’t want or intend to be critical here, was some of the crazy connections ancestry offered for members of my tree. Now generally they are useful but at this simple level I feel they could be a bit smarter. (I’m in IT development by the way so I know how difficult developing clever software using what is known as ‘fuzzy logic’ is. And IMHO Ancestry as an example of a website is very slick).
But before I get started I wondered if anyone had seen any research into this aspect of our hobby? Any references might save me reinventing the wheel.
Cheers Nigel
p.s. I know the detection is most of the fun and I don’t aim to look to automate that. I just am interested in generating probability of accuracy. For example an algorithm might examine a tree documented with sources and identify weaknesses that warrant further verification.
Cheers Nigel
Comment