Why no record hint for birth source?

Paul W

Without any marriage or death detail being attached to this ID, the hinting algorithm has produced correct suggestions for both these events. However, in spite of the birth detail being inputted, and there being a source to attach to match, this does not appear under Research Help. I have found this to be the case whilst working on siblings of this the individual, over the past couple of days.

Using the link, via the FamilySearch logo on the Person page, the birth source appears as the first item on the Results page, so why doesn't the Hints algorithm lead to it appearing on the Person page?

(Naturally, I will return to this profile later to complete the missing data and add more sources.)

Gordon Collett

I didn't look at the hints that were on his Family Tree page. Looking at those, I retract everything I said. I don't seen anything in those marriage or death hints that give any reason as to why they did get posted as hints.

Here is another theory for you: I do know that each collection has to be processed in some way then released into the Hint engine. Maybe that birth record collection has just not been dealt with yet?

dontiknowyou

The answer to that may lie in the relatives' profiles. If there are duplicate profiles or, alternatively, conflated profiles, the hints algorithm can be misdirected.

Paul W

@dontiknowyou

Thanks for the suggestion, but I don't think so in this case*. I believe it is more likely to relate to a flawed algorithm, as I have generally found it quite common not to have birth hints presented on the Person / Details page.

(* I checked both "Find Similar People" and "FIND" and he doesn't seem to have been confused with anyone of similar identity, or have a duplicate.)

A shame the engineers don't engage with us here, or it might be possible to get the issue checked out.

(BTW - I'm leaving things as they are until tomorrow, so others can have a look at the profile in its "raw state" - well, apart from the census sources that have been attached - to see if they have further ideas.)

Gordon Collett

Just a complete guess and idle speculation! The hinting routine is very strict. Early on it was quite a bit looser and people complained a lot about all the incorrect hints that others were just attaching without checking the records. They have made the algorithm so tight now that I can't remember the last time I ran across an incorrect hint. The trade off, of course, is all the correct records that don't appear as hints.

It is an unescapable fact of life and statistics: Never miss a true positive and you will have a lot of false positives. Never have a false positive and you will have a lot of false negatives.

In light of that, I think the hint routine puts a lot of weight on relationships. I suspect that in the illustrated case the fact that the record does not have any parents drops the probable match score too low for it to appear as a hint.

The search routine is weighted more to never missing a record which is why for every true positive in the search results we get hundreds of false positives.

dontiknowyou

Yes, the engineers did adjust the hints algorithm to err on the side of missing hints rather than offer so many bad ones. I think this was announced at RootsTech in 2020?

That said, the question is why in this instance is the algorithm missing this particular hint.

Gordon Collett

Because there are no parents in the birth record.

Would you want the hint routine to declare that there is one and only one Stanley Alexander Tungate born in all of Norwich about 1900?

Taking that birth record in isolation with no other supporting evidence that the birth record really is for Paul's Stanley is the classic "Same Name - Same Person" error that led to so many complaints of sources being incorrectly attached when the Hint routine was looser with its matches.

Paul W

I appreciate the comments / ideas (especially about the marriage record), but would have thought that at least the same criteria should be applied in the case of the death record - which is shown as a hint - as for the birth record, which isn't. The possibility of two individuals of the same name having been born, or died, at the same place (matching birth years) and time surely have the same odds?

Paul W

@Gordon Collett

I do know that each collection has to be processed in some way then released into the Hint engine. Maybe that birth record collection has just not been dealt with yet?

Yes, I think that's probably it. Well, short of a response from one of the engineers, I don't think we'll get any closer to the probable reason than your suggestion.

dontiknowyou

Maybe that birth record collection has just not been dealt with yet?

I am sure that's not it. I have consistently seen hints become available just as soon as the collections do. But this collection has no hints. I browsed Search results of 1000 birth registrations for a surname I've worked on using that collection, so I know there are Family Tree profiles matching records. As expected, some records were attached to profiles. But not one record had a hint icon on it.

My hypothesis is that this collection has been excluded from the hints system. I think the issue is that English birth registrations are quarterly and the year on the profile isn't close enough to generate a match. As a test of this hypothesis, try adding the birth date (from the death record) to the profile and wait a few minutes to see if the birth registration hint pops up.