Life Event exact search when searching for people doesn't work
I searched specifically for people who died in a certain area at a certain time (1826-1839) and it's pulling up records from the 1900's who are obviously not a match. I can't figure out how else to report this error than on here.
Answers
-
It is unexpected and undesired behavior, but I believe this is the result of a deliberate programming choice, not an error. The reason for the behavior is that the search parameters have a logical value of "or", not "and". In other words, filling in a place and a date range gives you everyone who matches that place, with any date, and everyone who matches the date range, with any place. The ones that match both should be higher up on the list; this means that if there are good matches to your search parameters, you may not even notice the or-not-and results down at the bottom. It's only when there are no really good matches that the "for some value of X that I am not familiar with" results are noticeable.
You can remove some of the undesired results from the list by using filters. These are applied after the search and have a logical value of "and": if you tell it to filter for death place = United States, Pennsylvania, Mercer and then for death date = 1800, 1830, your list should consist entirely of people who died in Mercer county in the 1830s.
Except of course it doesn't. There's still Delaware and Somerset and Perry county results in there. I think this must be the result of some error or irregularity in the Places database: its definition of "Mercer county" is too broad. Or something.
0 -
I'm failing to see what I am doing differently, yet I am only getting 140 results compared to the 5,542 figure shown in your screenshot.
That said, no - the "Exact" (only) results option, many of us argued for a long time to mean produce precisely that, seems to have vanished - judging by the inclusion of results including records of Ohio and Indiana in the list.
There is a problem in that FamilySearch employees very rarely visit this Community forum to address such questions. Specifically, has this new behaviour produced by accident or design? Perhaps a moderator would be kind enough to escalate the issue for investigation by (and hopefully a response from) the engineers.
0 -
@Paul W, from the looks of the screen shots, @east1 is using Find in Family Tree while you used Search under the historical records.
Here is the search result URL: https://www.familysearch.org/search/tree/results?q.deathLikeDate.from=1826&q.deathLikeDate.to=1839&q.deathLikePlace=Mercer%2C%20pennsylvania&q.deathLikePlace.exact=on&q.givenName=Josephine
Basically, I would agree that the search engine is designed purposely to give these results. This falls into the category I call "an overly helpful computer." What you were probably expecting to see was a "no results found" message because there are no Josephines in Family Tree who died in Mercer between 1826 and 1839. Otherwise she would have appeared. But the programmers just couldn't bear to leave you empty handed. I find that Ancestry and My Heritage take the same philosophy of "maybe the user doesn't really know what he is looking for" and throws in a bunch of results based on just one of the criteria, as Julia explained.
One search technique I have found particularly helpful in Family Tree, especially when using an exact place name search, is to spell out the complete place name as you would expect it to be standardized in Family Tree. There is no such place as "Mercer, Pennsylvania" as a standard in Family Tree. In this case, modifying the search criteria to have Mercer, Pennsylvania, United States, however does not make any difference because there are just no Josephines that fits.
Something else I have noticed, is that when there are no results and the bonus possibilities are thrown in, the date range take priority which is why the Josephine who died in 1837 (in range) in Delaware county (wrong county) is listed before the Josephine who died in 1974 (out of range) in Mercer county (correct county) came first.
Elizabeth Josephine who died in 1974 (way out of range) and Martha Josephine who died before 1893 (closer to the range but still quite a ways outside it), both in Mercer, likely have the same ranking and are just listed alphabetically.
I have gotten in the habit, which is probably a bad habit, here in Family Search, and in Ancestry and My Heritage, to assume that if my expected results are not on the first page of results, particularly if the very first result is way off, that there are no matches for my criteria and try a different search.
It can be instructive to play around with the criteria and see how it affects the results. For example, I modified your search to have the full place name but decided to make sure there was a problem with the spelling of her name, so used Josep* to catch Josephine, Josepine, Josephina, Josephinia, and any other possibilities. This gave these results: https://www.familysearch.org/search/tree/results?count=20&q.deathLikeDate.from=1826&q.deathLikeDate.to=1839&q.deathLikePlace=Mercer%2C%20Pennsylvania%2C%20united%20states&q.givenName=Josep%2A
Here, the first seven results match the criteria perfectly. Then you get a bunch where the dates are in range but the places start falling apart, although they are all in Pennsylvania on the first page. But there is only one Josephine in a sea of Josephs, the one who was first in your results list. Confirming again that there are just no Jospehinas in Family Tree who died in Mercer from 1826 to 1839. If you know there is one and and have seen her in Family Tree, then you need to post her ID number so that her record can be examined to figure out why she is not showing up in the search results.
1 -
Thank you for pointing out what I had missed - FIND v. Search obviously producing different results (in numbers and otherwise).
On the general issue, regular users have come to accept that FIND only prioritises on "Exact" matches, but there does appear to have been a recent change in the Search algorithm, whereby that is now far less likely to produce a list of "exact" results, after us checking the boxes.
As already stated, I wish we could find out more easily whether the engineers are changing things by accident or design.
0 -
To complete my essay, let's take a look at Search. Because place names are handled a bit differently, wildcards are a good idea in Search so I'll search on Josephine, died in Mercer*, setting the exact match flag, between 1826 and 1839. I will also limit my search to just death records and just records from Pennsylvania: https://www.familysearch.org/search/record/results?count=20&f.recordType=2&q.deathLikeDate.from=1826&q.deathLikeDate.to=1839&q.deathLikePlace=Mercer%2A&q.deathLikePlace.exact=on&q.givenName=Josephine&q.recordCountry=United%20States&q.recordSubcountry=United%20States%2CPennsylvania
Here again it appears there are no "real" results, just the bonus "maybe these are of interest?" results.
But the rankings on the first page are interesting. First come the only three Josephines, none of which have a death or burial date at all. Then come just Josephs with patterns suggestive of how dates and places are handled and ranked.
It is very interesting that again Delaware is viewed as a good match for Mercer. I wonder if the fact that there is a Delaware township in Mercer county causes those to be pulled into the results.
0 -
I agree with you on two things here. Firstly, baffled by the Delaware (township) results being included so high in the list, I checked it out on Google Maps and - due to its close proximity to Mercer (actually within Mercer County, you confirm) - I too decided this was the reason for its inclusion in the results.
Secondly, on the wider point of wildcards, I also searched on Mercer*. That wildcard can make a definite difference to the number of results. In other searches, I frequently get no results at all if I just search on the name of a town / city, but get returned all those results I was anticipating (and probably more!) just by adding that wildcard.
0 -
The Josephine being returned for "Mercer" died in Delaware county, just outside Philadelphia. Mercer county is nearly the furthest away from that that you can get while staying in Pennsylvania: it's just across the state line from Youngstown, Ohio. (Crawford and Erie counties are further away from Delaware county, but not by a whole lot.)
Given that the death location of Delaware county is fully and correctly standardized, its return as a match for Mercer is nonsensical and erroneous. (No, there is no Mercer township in Delaware county.) The definition of "Mercer, Pennsylvania" must be incorrect in the database.
0 -
I don't see anything in the Places databases that could cause any confusion between Mercer county and Delaware county so either there is a flaw in the Find routine or something else is going on.
I suspect the real issue is that because there are no full matches at all, the place name is being moved up one level, that is the search engine drops back to searching Pennsylvania, rather than Mercer, Pennsylvania, and that there really is only one Josephine in the databases who died in Pennsylvania between 1826 and 1839 so when the date is taken as the second ranking value after name, and the exact match flag is used against the truncated place name, that Josephine in Delaware county is the best match.
Or maybe not. Removing Mercer and just searching Pennsylvania, United States does bring up a full first page of appropriate results except for two that look goofy at first glance: https://www.familysearch.org/search/tree/results?count=20&q.deathLikeDate.from=1826&q.deathLikeDate.to=1839&q.deathLikePlace=Pennsylvania%2C%20united%20states&q.deathLikePlace.exact=on&q.givenName=Josephine
Of the two that look out of place, one has a non-standardized death date and the other has a death date of 1936 but a burial date of 1836.
Long ago the search results page had this nice, blue diving line labeled something like "The following results do not match your search criteria but may be of interest." I wish they would bring that back. I'm pretty sure that it would be the first entry on the search being considered here.
0