A number of problems in familysearch lately over the last week

LegacyUser

Justin Masters said: I've noticed a couple of problems in FamilySearch, two today, and one going back a week or more.

I've noticed the hints is not offering as much anymore for hints. I'm not sure why, except that perhaps they may have lived in the last 110 years (even if they're dead). I can't find rhyme or reason for this, and it's more of a perception or gut feel, given the amount of source hints that I can dredge up myself with simple search terms. And once I even spelled out a name as it showed up in the indexed entry, and it still refused to find it. (I had to try two other people in the family before I found one that actually worked. (sorry, I don't have the persons, but likely the 1940 census found here: https://www.familysearch.org/ark:/619... ) I found it easily at ancestry, but had a very difficult time getting it found in familysearch.

Tonight, several people in which I tried to attach a source, I had to click TWICE on a living or deceased button to get it to work. The first time shows the radio button fading, with a white box around the border of the radio buton control, and the second press caused the color to fill in.

Uploading a photo for the memories doesn't seem to work, resulting in a blank image, and trying to upload it again highlights that file as a duplicate which is then subsequently deleted. This is found here: (hmm... well, now it shows up, but initially it did not for a long time). https://www.familysearch.org/tree/per...

LegacyUser

Lundgren said: Separate topics for your issues will probably make this easier to discuss. This thread, if it becomes active, will go all over the place very quickly.

With regard to searching the historic record:

https://www.familysearch.org/ark:/619...

Using the following searches:

She is the first result in a simple name search: https://www.familysearch.org/search/r...

Adding a birth range, she is still first: https://www.familysearch.org/search/r...

Adding a birth place she is still first: https://www.familysearch.org/search/r...

If you can provide some sample URLs of your searches that you used to try to find the record, then we can see if we can understand what is going wrong and perhaps how to fix it.

LegacyUser

Justin Masters said: Hmmm... that wasn't my experience when searching for her a few days back.

I tried to reduce the collections to the 1940 Census, couldn't find her, and then thought it was silly that it wasn't found (not that people aren't skipped during census counts, or that records are badly transcribed), but that's what led me to look on Ancestry, and easily found it there.

I checked to see if perhaps the name could be badly written and names being misinterpreted or mis-indexed, and didn't see any candidates like that.

I came back to familysearch and tried a few different people in the family and found no results, and was getting really frustrated because it was RIGHT THERE on Ancestry. And finally a hit! But... it was a bit disconcerting.

In addition, I was surprised that there weren't hints for them in the research help portion of the page. (and I've seen the lack of hints there more frequent over the last week or two, leading me to wonder if there were some policy reason being implemented.)

The findagrave hints seem to be better, but others seem worse.

and I'm sorry, I can't point to particular searches... it just "seems" like fewer research hints are being supplied, when they can be found without too much effort.

Thank you for looking into my issue.. .and good point about not combining issues.

I hope someone will look into whether or not the algorithm for providing recommendations has been changed without adequate testing (by testing, I mean, comparing returned Research Help hints for the old algorithms (on a number/quality basis) against the new levels (number & quality) hints across a large data set.)

LegacyUser

Lundgren said: The system that provides hints and and the search system are completely separate systems that use different technologies. I can't speak to what you are seeing with hinting.

In the future if you run across cases like this for search, please provide the searches you are using (just copy and paste your URL) as well a a link to the record you are after and we will see what we can find.

Thank you for your feedback!

LegacyUser

Justin Masters said: Thanks for clarifying the separation of the hints and search systems. That makes sense.

I'll try to remember the searches. I just kinda "lived" with them, until I had other issues pop up, and then threw them all into a single email.

Sensing that you're focused on the search engine part of it, I want to hopefully clarify that the Record Help portion seems to be providing less hints/suggestions, and while not clearly separated in my original post, is an issue (too strong a word - problem?) that seems to have crept in. Again, it's a perception thing... but if you'd see the kind of stats regarding my efforts last year in family history work and attaching sources, etc, you'd see a LOT of work, and hopefully it's not dismissed casually. I *AM* seeing fewer Research Help suggestions... and I don't have a way of quantifying it, other than to question why certain record helps aren't being displayed, but can be easily seen with a search or two. (The number of census hints have gone down a bit, surprisingly!)

I don't have a background in statistics, and I'm trying to avoid any self-perceived biases. And I don't have the capability to do a large dataset comparison with (possibly?) old vs new algorithms.

Just trying to be helpful.

Thanks again!

LegacyUser

Lundgren said: The research suggestions are also not part of the search system, so someone else would need to look into that...

If you've got specific searches that aren't working please send along a link. Even better if they aren't returning a specific record you know exists.

Thank you again for your comments.

LegacyUser

Justin Masters said: Lundgren,

Here is an example of not only record hints not working, but the search not returning a good result that should be happening, even with filters applied to collections.

The person is Robert Lee Williams GQLS-4RD
I'm looking for him in the 1910 Census (or 1900, but for now, I've been looking for the 1910 census where I see this issue popping up.)

I won't attach this, so that you can see for yourself.
There's NO record hint provided for the 1910 census.

A search of familysearch using the FamilySearch link on his page results in this query: https://www.familysearch.org/search/r...

The 1910 Census for him does NOT appear on the first 300 results (3 pages)

Clicking on Collections and picking 1910 ONLY, I get 12 results, of which he is NONE of these.

So, now, let's go over to his brother Carl.

There's a record hint for him in the 1910 Census which points to
https://www.familysearch.org/ark:/619...

You'll notice that the people in the family structure are correct, with proper spellings and birth places (I'm NOT touching any people in this family, because I want you to see this)



If I right click on Robert's name on this popup hint found in Carl's page, it points to:
https://www.familysearch.org/ark:/619...

Looking at the record vs family structure at https://www.familysearch.org/search/l...

Robert is not attached to anybody else, nor is he pointing to anybody else.

When I go back to Robert's name in Carl's hinted 1910 census record and go to the link above ending in MKXX-DCK, and then click Review, I go to:
https://www.familysearch.org/search/l...

and oddly, it's asking me who I should be attaching this to.



Hopefully that helps you to debug why Robert's not receiving a hint, nor showing up in the search results or filtered collection results.

Even forcing a refresh on Robert's page after doing these steps above doesn't show a hint for Robert.

I'm using Firefox (latest version) on Windows 7 x64.

LegacyUser

Brian Jensen said: Justin,

Thanks for your response and detailed investigation. With just a couple search attempts I was able to find Robert Lee Williams in the 1910 census:

In this query he is the first hit:
xhttps://www.familysearch.org/search/r...

For our Search system I will often start with minimal info and keep adding more details till I find the person I’m looking for. I started by searching just the 1910 census with Robert Williams, Birth Missouri 1900 but I didn’t see him in my responses. I then added mother’s given name: “Christina” and your Robert Williams was the first hit.

Regarding Record Hint:
I will need to do some more investigation why we didn’t return a Record Hint for him but my guess is that there are several factors that reduced his match score:
1) different birth years (actual record is 1899 but census is 1900),
2) common names (Robert and Williams are very common names and our classifier takes that into account),
3) Father’s given name is not an exact match (Edward C vs E C),
4) Given name is not an exact match (Robert vs Robert Lee).
These all probably resulted in his match score being below our high confident match threshold. And our goal is to only show high confident Record Hints.

I also looked for this family in the 1900 Census, but I’ve had no luck finding them with FamilySearch or Ancestry. But I’ll look some more next week.

Thanks,

Brian Jensen
Search and Match Engineering Manager

LegacyUser

Lundgren said: I've only done a quick look at this so far.

The link to search historical records should be considered a starting place not the final search or the best search for all records. It is a compromise search that is far from perfect.

I've modified the search from the tree person to make it more relevant for the record you are interested in.

I removed the death info as that isn't relevant for a census. I filtered the search to the 1910 census. He was a minor in 1910. Since I restricted the search to the 1910 census I was able to add his role in record to "son". I also added his father's given name.

The following is the search:
https://www.familysearch.org/search/r...

I believe your record is the forth result.

Please let me know if this isn't clear.

The hinting system does not find every record for every person in the tree, and it only shows the best record hints to avoid false positives.

LegacyUser

Justin Masters said: Thaks Brian.

I'm going to separate the Hints vs the Search part here:

===============
SEARCH
One of the points I'm trying to bring up is that WITH that nifty little link on Robert's page, it auto-populates info for the search, and the use of that "help" shouldn't detract much from what is found. However this is something else going on under the hood (invisibly) that affects the search results.

(I do agree with blank searches, I start minimally)

The differences (as you point out) are minimal.

What I saw in my analysis of the returned result after using that FamilySearch search link on Robert's page was that the results coming back selected military records, marriage records, etc. but even the few census records (buried down aways) didn't return the info. And as I pointed out, the filtering using the collections filter (1910 census) didn't return him either. So something else is going on here...

To emulate Carl's search with Robert, I deleted all death info after clicking the FamilySearch Search link on Robert's page. I'll point out that Carl's birth info is more vague, with just a year and a state, while Robert's is highly specific.

With Robert's death info deleted from the pre-filled out search form, I STILL get the following clusters of record results in this order:
- a death record (with birth info as well - a match),
- veterans and draft records,
- marriage records,
- GenealogyBank records,
- Findagrave records,
- a smattering of assorted records, including a 1940 census record and a wrong 1910 census record result.
- more draft records
- more census records, draft records (and still not found within the 1st 200 records)

When I filter by collections for the 1910 census (again, death info deleted), I get 168 results back (vs the 12 I got back with the death info included - gotta wonder why on that!)

I do find him about 20 or so entries down.

So... either it's taking OTHER vital (or previously attached records') info invisibly into account in the results (with the death purposefully deleted), or the functionality significantly changes with a more specific date and place of death.

===================

HINTS

For hints not showing up, it seems to disregard scoring that SHOULD occur when it takes into account not only the relationship between the individual and the parents, but perhaps the known siblings and those found in the records.

Again, this seems to be significantly affected by either the inclusion of the death info for Robert, or changes significantly with the much more specific birth info.

(My money is on the latter, and would probably make better sense with my perception (mentioned above) that I was seeing fewer hints in the last few weeks.)

If I first enter specific info in the course of my doing work on obituaries (often mentioned in newspaper obits as born _____ in ___ to ____ and ____ ), and get the more specific record hints (birth, death, SSDI, findagrave, GenealogyBank obits, military records, etc) but NOT the census records hints, then.... uh...
it's not helpful for finding additional family members through censuses.

So that should probably be looked at.... more specific info should NOT lower the hints rankings (or search results) for census records.

Hopefully that's not too confusing... I realize I'm asking you to follow my train of thought, which may or may not be what your algorithms are meant to do.

But my way of thinking makes better sense. :-)

LegacyUser

Justin Masters said: Thanks Lundgren.

It does seem that adding the position/role of "son" by naming the father seems to help.

But... that seems to over-complicate a search by having to do that, and seems to dismiss (in the initial search) the results that SHOULD come from a census collection filter, regardless of whether or not death information is added to the initial search criteria.

If anything, death should be considered in the search to exclude censuses AFTER his death (in this case, he died in 1937, and no 1940 census records should be returned).

And yes, I realize that's a logical statement, and not an algorithmic one.

LegacyUser

Lundgren said: Some collections have special fields. The 1910 census is one of these collections.

Once you filter down to that collection only a new friend shows up on the search paraments. In this collection, it the "Relationship to Head of Household" field.

That is where I specified "Son". Only some collections have these fields, so they are not offered on the general search page.

LegacyUser

Justin Masters said: Lundgren, I have a number of questions of what you just said, so please, don't be defensive. I'm just trying to understand.

Could you explain what special fields are (regardless of which census), and why they matter on the 1910 census search results, particularly if the special fields are not "reachable" via search criteria?

I'm not understanding "a new friend" in your sentence. Was that an auto-correct? (I'm trying to understand what other word could fit there and still make sense, so I'm probably missing some context.)

LegacyUser

DougHo said: Brian - regarding #1, a pet peeve of mine is that FamilySearch usually does age "wrong" for US census. Someone age 10 in 1910 census is more likely to have been born 1899, not 1900. Someone born in 1900 would only be age 10 in 1910 census if they were born Jan 1-Apr 15 (around 29% of the year). Otherwise to be age 10 in census with birth date Apr 16-Dec 31st (around 71% of the year) their birth year would be 1899.

So since Robert L was born Nov, age 10 means he was born 1899.

LegacyUser

Justin Masters said: It's not entirely Family Search's fault. It goes based on what was answered, but also the question. (I know you're scratching your head, so let me share a rather eye-opening instance that further muddied the water.)

I was searching for someone who was born in early 1920 and SHOULD have been on the 1920 census. But someone much more knowledgeable than I (and more observant, since I didn't notice this) pointed out that the 1920 census was given right near the beginning of the year and asked about who was alive at that time).

Okay, you're saying... soo?

Well, it's the only census that was done based on the beginning of the year, when all others were done in the middle-ish part of the year. (I'm sure trying to reach everyone in the snow during the winter time presented some fresh answers to census taking logistics, hence the return to mid-year census taking later.)

But the question and when it was asked affected a Census output.

This doesn't even go into whether the people were honest in what they told the census takers, or what the census takers heard or wrote down.

Now, the 1900 census DOES have a question that says "Age at last birthday" and has a month and year of birth. And I'm not sure if FamilySearch actually calculates based on the month and year, or if it just uses the answer from the age column.

LegacyUser

Lundgren said: I didn't mean to come across as defensive. Mood can be difficult to convey in text. Sorry about that.

" only a new friend" was auto correct... It should have read "a new field."

Some collections have fields indexed that are available only to that connection, or a subset of connections. The 1910 collection has a relation to head of household field. Other collection might have race. These fields can be seen only in collection specific searches, either from the collection page or once you have filtered into a specific collection.

If you go to the 1910 collection page, https://www.familysearch.org/search/c... you will see the added field.

In the 1940 collection specific page, you will see the residence in 1935 field, but not the relationship to head field: https://www.familysearch.org/search/c...

The 1930 has relation to head, but for obvious reasons, it does not have residence in 1935. https://www.familysearch.org/search/c...

LegacyUser

Justin Masters said: Lundgren, thanks for your follow-up.

I looked at the first link you provided for the 1910 census, and maybe I'm missing some nuance suggesting the use/existence of the special fields, as the search looks exactly like what you see currently in the normal search of historical records.

I did see a breakdown of the 1910 census fields (not necessarily indexed) by following the "Learn More" link.

And I'll admit to some "sloppiness" in my assumption that relationships have been shown since the 1880 census.

But here's where I'm puzzled... the 1940 and 1930 Censuses DO have relationship to head of household info (follow the Learn more link seen on each of those links you provided)

Prolly a typo, but that would be COOL if 1935 census info could be found on a 1930 census! :-)

With regards to the 1935 census info you mentioned (if you don't mind my asking), why DOES the 1935 census location not get recorded in the 1940 census records (and this is confusingly implemented in the attachment of records to people in the family tree). If they continued to live there previously, then "same" COULD populate the same info as the 1940 census field, with the exception of people obviously not born yet in 1935.

Was that field just not recorded when indexing the 1940 census? If so, how come it's included in a 1940 census field attachment screen? (I reported a bug on this previously, for the funky values that got added in here, and it was at least fixed so that it could be modified to be filled out manually or made empty (YAY!))

LegacyUser

A van Helsdingen said: On the theme of issues with the website over the last week or so, I have noticed:

1. Often getting error messages saying an Internal Error has occured, especially when using the Catalog.

2. I am unable to index, since I always get the error message in the attached image when trying to open a batch.

3. The US Federal Censuses have had restrictions applied to their access that apparently were not intended- there are now accessible only at a FHC or Affiliate Library, see: https://www.wikitree.com/g2g/998075/f...

4. The "What does your last name say about you" feature doesn't work for surnames of more than one word, like my own surname "van Helsdingen", which apparently has only 3 people in the FSFT (A search directly in the FSFT reveals many more, and I personally know I have added at least 10 people). But "van Elsdingen", a rarer variant of my name, is said to have over 3.4 million people in the FSFT. Only 45 people were recorded as having that name in the Netherlands during 2007, and all descend from a man born in 1808 (the brother of my ancestor).

The feature gives reasonable and correct-looking results for surnames of one word only.

LegacyUser

David Newton said: Perhaps it's the copyright holder that's restricted the US federal censuses. I mean it's not like they're going to tell us is it since that's settled policy and apparently written into their contracts.

Oh what was that? Yes that's right the US federal censuses DO NOT HAVE A COPYRIGHT HOLDER! So what about it Familysearch? What is going on with these restrictions? Don't try and hide behind "contract terms" in this case. Just give a proper explanation for what is going on.

Incidentally perhaps the restrictions like on the Scottish censuses and the English 1939 register should be recorded some place like Wikitree or We Relate. That's a place we can direct people to for a proper answer to the questions which Familysearch can't censor.

LegacyUser

Douglas Burdette said: "I had to click TWICE on a living or deceased button to get it to work. The first time shows the radio button fading, with a white box around the border of the radio buton control, and the second press caused the color to fill in."

Past this, when the record is created, it's set to "Living", no matter which option you choose.

This is a known issue. They're looking into it.

LegacyUser

A van Helsdingen said: FS Support has told me that there are indeed projects (e.g. UK non-conformist records) that have restrictions on who can index them.

Why would FS and/or a record custodian do that? Do they want people to help or not? Do they really think people are going to index in order to access images they otherwise could not access, rather than with genuine motivations.

So I've asked Support if my non-LDS status, or geographical location (New Zealand) has anything to do with why I no longer have rights to index any project. I should also have added I am doing this from a home computer.

LegacyUser

A van Helsdingen said: Following another message from Support, I have found that this was just a technical problem, either with the FS website or my computer. While it is true that some projects are "Closed" to certain groups, this did not affect the projects I was seeking to index.

LegacyUser

David Newton said: "While it is true that some projects are "Closed" to certain groups, this did not affect the projects I was seeking to index."

I believe that the appropriate phrase to describe that state of affairs is completely nuts! As you say what on earth do the record custodians think they are doing? They are limiting their potential pool of indexers and thus deliberately slowing down the indexing process.