Knowledge of 17th, 18th 19th Century life
Your algorithms seem to lack any awareness of life in earlier centuries. They need to be monitored or corrected by someone with such knowledge. Few people lived in cities.
There are good resources (such as Everton) with a record of when Counties were formed. They need to be used.
When a Census record is attached as a source, does it also need to be "tagged", whatever that means?
項留言
-
Yes, a census should also be tagged to the residence. Until very recently, that was not an option on the FSFT. Now that it is an option, it's very easy to do, manually, on existing sources. Or, using the source linker, it happens as a matter of course when attaching the source to the profile.
3 -
The algorithms do take into account time and place. That is why the Data Quality Score is only available for certain places and certain time periods. The programmers have been very receptive of information from experienced users who can provide the information needed to fine tune the algorithms. However, they need specific information, not vague generalizations.
You should compile a list of specific items you find that need to be adjusted and post them here. Posting the ID of a profile with each issue will help them out a lot.
Regarding the difference between attaching sources and tagging sources, sources are attached to the entire profile and then appear on the profile's Sources page. When attaching a source or by going back after a source is attached, some specific data items can be tagged with the source and the source will then appear in the Data View popup for that specific data:
1 -
Keep in mind that certain generalizations will always need to be made and each person's specific situation in life will never be able to be accounted for in any algorithm. That is why we can dismiss the quality check items. I'll admit that it is not obvious that we can since there is nothing that indicates that clicking on the item causes anything to happen, but it does:
3 -
If the algorithms were taking into account the century, they wouldn't be demanding a city for every event.
0 -
The point is that the algorithms can take into account the century, not that they are currently doing so for all possible issues. To improve the algorithms the programmers need information from us users as to what needs to be added and why. If you work in areas where there will never be more than a three part place name, they need to know specifics of when and where.
You can see in the Algorithm Update post at the top of this group changes that have been made based on user input. Unfortunately that has not updated that list since October.
2 -
You may want to review this thread:
where the final post reports that this issue has been passed on to the programmers.
1 -
Oh, and it probably wouldn't hurt to mention here even though it is a separate issue that this particular quality flag is badly worded because what the routine is looking for is a four level place name. It doesn't know and doesn't care if that fourth level is a city or not. I mainly work in Norway in Family Tree and the vast majority of my wife's family there lived on farms. The quality checker is quite content with place names of the format Farm, Community, County, Country.
1 -
A full and complete placename in Hungary, both historical and modern, consists of a grand total of three levels: place, county, country. It's true for Budapest on down to the tiniest village. Or actually, I take that back — Budapest after 1950 has just two levels: Budapest, Hungary. When they reorganized the counties, they made the capital city its own, separate jurisdiction, with nothing above it. You can get a fourth level by adding extra detail, such as an outlying farm name or a street address, but that's not what the rating message talks about.
I think someone in the Places team has tried or is trying to "remedy" this by adding districts between the place and the county, but I believe this to be a badly misguided decision. In some counties, districts changed every few years — borders, names, administrative centers — and nowhere in the country were they actually necessary to uniquely identify a place, especially not after what I call the Great Placename Disambiguation Project of the very late 1890s. Adding districts to all Hungarian places would be like trying to add electoral districts to all U.S. placenames: an utterly nonsensical mess.
All that said, I have only encountered the "no city" message once so far (https://www.familysearch.org/tree/person/details/LR9T-BMQ). I somehow managed to dismiss it without an actual expletive. Ah: one of his sisters and both of his brothers have it, too, I just haven't previously mustered the fortitude to look at them.
0 -
@Gordon Collett said
… Oh, and it probably wouldn't hurt to mention here even though it is a separate issue that this particular quality flag is badly worded because what the routine is looking for is a four level place name. It doesn't know and doesn't care if that fourth level is a city or not. …
I beg to differ. I think! In the thread "Data Completeness 'Place missing a city" is inaccurate' that you linked, I came to the conclusion that
A burial place of "Copenhagen, Copenhagen, Denmark" (i.e. 3 parts) did not generate the "missing a city" message. The same profile's death place of "Frederiksberg, Copenhagen, Denmark" did generate the "missing a city" message. So the same number of parts but different messages.
When I looked a little closer at the 2 places though, light began to dawn - the double Copenhagen is a city in the Standard Places. Frederiksberg, however, is a Lutheran parish. (The city of that name is "Frederiksberg, Frederiksberg, Copenhagen, Denmark"). So the death place of "Frederiksberg, Copenhagen, Denmark" generated the "missing a city" message because it's literally that - it's got a parish, not a city…
From a programming point of view - given that some countries have standard 3 part names, and some 4 part, while "England" moved from 3 to 4 when it became part of the UK - I think it would be both easier and more robust to look at the type of the place. Obviously I hope the check is on a more generic value than "city", but it should allow village, town, etc.
0 -
@Adrian Bruce1 Interesting to hear the conclusions of your investigations. So the quality check routines must be quite complex and probably getting more complex with every request for fine tuning. As I mentioned, none of the Norwegian places I usually work with have any city at all and the quality check routine is fine with that. But it may be treating the community as a city even though the communities are more like counties than cities.
1 -
@Gordon Collett said about "his" Norwegian places:
… it may be treating the community as a city even though the communities are more like counties than cities. …
That is probably the 64,000 dollar question - what level of "place" counts as precise enough to satisfy the "missing a xxxx" check? And should that level vary across the globe? (Bearing in mind such traps as the difference between a typical "city" in the US and one in the UK).
2 -
The algorithm doesn't accept a townland in Ireland as qualifying. And, it still calls out rural places in 19th-century US as missing a city.
1 -
… and since townlands in Ireland are smaller than some settlements elsewhere, then I'd suggest that they ought to be precise enough. That's assuming that they are marked as such, of course.
1 -
Yep, things are pretty complex.
Here I put in a place name with just province and country and got that no city message:
I added a farm name and the message went away even though I did not add a city:
and even though this is a very low quality place name in my opinion. Oppland province is divided into a couple of dozen communities and not knowing which community Skjerdalen is in is a big flaw (This is slowly being rectified in the Places database but it will take decades.), particularly since there could be a dozen Sjerdalen in Oppland. But having this reflected in the data quality checker would be problematic since the Places database's standards has hundreds of thousands of these incomplete place names for farms. And since the quality checker looks at the standard and not the displayed text, as far as I have seen, they really can't require the full place name since there is no standard for so many of these place names.
I don't envy the job of those programers who are working on this.
1 -
But to get back to the point of this post and the original question, @Mormor192 , can you post the ID number of a profile in FamilyTree which has a place name with this "missing city" flag that is fully and correctly entered with the maximum amount of detail that it ever can or ever will have and explain why that is and always will be the complete place name so that the programmers can address your concerns.
I'm sure we'd all like to see if this is a Places database problem or a Data Quality Check problem or something else. I think we have already determined that the use of "city" in this flag is a bit deceiving because the checker views Norwegian farms, among other things, as "cities."
2 -
I'll hunt for some more when I have time. I've already dismissed, with or without a reason, a great many for people in West Virginia and Virginea who lived on farms, not in towns or cities. The farms don't have names, so it is County, State, and United States or British Colonial America.
0 -
Having taken a closer look at the "missing city" profiles I've encountered, I think the algorithm is using the place type assigned in the Places database. Problems arise when there are two entries in the database that are identical except for their type, and only one of them is "acceptable" as a city.
(My solution for this particular case would be to merge the district into the municipality. As I wrote above, there is no practical purpose for having the district in the database.)
1 -
@Julia Szent-Györgyi - I agree with you on the Place Type. I hadn't thought about the "identical apart from the type" issue. There must be a considerable risk that some profiles use the district (which presumably generates the warning) instead of the desired town / city etc. This situation is, I think, made worse by the fact that I have never found a key combination to tell me the Place Type of the place actually on the profile - from memory, I have to update it to get the drop down that includes the Place Type. In such a case of confusion, it might be impossible to understand why the issue is generated in the first place or even what fixed it.
0 -
I ran across another example where place type makes a difference for the quality checker.
First some historical background. On 1 January 1965 Norway had a major rearrangement of its municipalities with boundary changes, splits, and merges. Many of the older municipalities that vanished on that day are not yet in the Places database and many of them that are in the database were given the place type of Division.
What I discovered is that municipalities correctly given the place type of Municipality do not trigger the missing city warning and municipalities incorrectly given the place type of Division do.
1 -
Gordon, you wanted an example of an ID where place is complete without a city name. I've dismissed dozens, the first ones with explanations on every item, late ones without as I gor tired of tyoing the same thing multiple times. Here is an example I haven't "dismissed" ID # LH15-BZQ. Born on a farm; married at the bride's home, in a country church or more likely at that time by a circuit rider (Minister who traveled from place to place by horseback); died on his farm and probably buried there in a corner of the property. Farms usually didn't have names in that part of the country. It was a frontier region without cities, churches, hospitals, etc. Many people fall into that category.
There seems to be little rhyme or reason to the determination of the quality level.
It would be more worthwhile for the people working on this at headquarters to counsel the people who are determined to connect themselves to a line that isn't theirs. To do so they change dates and places of births, marriages, deaths; remove valid sources that interfere with their desired line; insult contributors trying to help them. Several people before me have tried to explain and correct the data for one of these people, without success. All have given up. I was told "just don't upset the person". As a result hundreds of families have their lineage corrupted.
0 -
That's a good example. I see that the census records don't even record anything smaller than the county such as the township. And, ouch, you have a lot of flags to dismiss. I hope the programmers can examine that profile and come up with a way to deal with such. The more specific examples they get, the more they can improve the algorithm.
0
