โจ๐ IMPORTANT UPDATE ๐โจ
Comments
-
We really need to review the occupations. I would guess that at least half of the records I've reviewed had errors with the occupations. I have often used occupations as a search option to narrow down families. Please bring back those few columns.
3 -
I agree that the occupations are often incorrect - certainly more often than not.
3 -
@Karen Blackmore and @John Empoliti Have you registered your concerns by using the Feedback function?
1 -
I have not.
0 -
@John Empoliti Please, report this as Feedback. That is how it will get to the folks who can make changes.
1 -
OK. Will do.
2 -
Hi, I am new here, so forgive me if my questions sound silly but I would like to know what is AI what it stands for, and where do I find 'Family Review'? Is this different from Reviewing indexed record.
0 -
AI = Automatic Indexing or Artificial Intelligence, aka computerized (machine) indexed
'Family Review' is PART of the 1950 Census indexing see Get Involved page
0 -
Thank you Sam Sulser for the good news. The faster Family Review process will be much more effective. We don't need to debate whether occupations need to be indexed. Staying focused on names and relationships is a better use of resources.
2 -
It is truly unfortunate that people are making decisions to stay away from an amazing resource based on a process that they don't completely understand.
Please note - some of the fields that are not being reviewed by our volunteers *are* being reviewed by another group. There are a handful of fields where the quality was not at 98%. The fields that are not at that high rate are being reviewed through another process. Occupations is one of those fields.
Speaking of the concepts of AI indexes in general, if you could have an index of 40 million records, and 20 million of those were accurate, and 20 million of those were suspect, would you want to hold back the 20 million that were good because of the 20 million that had inaccuracies? What if it was easy to check and correct those additional 20 million records, but only if they were public?
What if we had ignored Wikipedia for the first 10 years of its existence because of its inaccuracies?
This is the model that FamilySearch is going to with its record creation, the ability to create automated indexes and then have a community of people who help and contribute to making them better. It will likely take years before we have the complete model working the way we'd like it to so what you're seeing are imperfect first steps.
FamilySearch has some brilliant people working on these problems. If you don't understand something, or if something doesn't make sense, please feel free to ask questions. There are a lot of pieces to this puzzle and if you don't see them all it's understandable that it would be confusing.
If these are the things we are hearing in the general public, we need everyone we can to understand the processes and be ambassadors to this work because we need everyone's help!
5 -
As I delve further into the 1950 records, I'm noticing the original Headers were often wrong. When I compare them to the ED maps or descriptions, I can see where the census taker left the city name blank or literally wrote "incorporated" on the Incorporated Place line of every page in the district.
3 -
I continue to be frustrated by the instruction to "uncheck" or "check" any line that does not belong - or does belong - in a Household identified by AI. When I make this correction then the Family Count is altered and the invariably the program moves me on to another review and the last family on the sheet is not reviewed.
The same thing happens if I realize an error and try to go back to correct.
How do we make corrections? What happens to those families that have been skipped by my review because the program moved me on?
Thanks.
3 -
I feel the same as many do, the translations have been horrible. Worse yet, the "short cuts" made now so that these can be accessed faster will be no help in the future for those of the younger generation whom are unable to read cursive. This will detour them to finding ancestors or stop them altogether in the search for their family. Just because we can look at it and read it, doesn't mean the kids that now can read it (most can't even write it). So much history is being lost by pushing this though. โน๏ธ
I feel to have rushed this project is a waste of time doing this project. To say you want to save it but then not have it to be usable for future generation doesn't make sense to me. ๐
3 -
@seahurstpeg I can understand your worry about getting things correct. Let me assure you that the system carefully tracks what has been done on each page. If a family hasn't been reviewed on a page, it will be presented to someone to review it.
In addition, before we attempted publishing content indexed by a computer, we put a correction process in place. This process has been available for a while on many of our indexed projects and can be seen as you search records. This way as people are working with the records, they can make corrections as they find them.
Sam ๐
2 -
@TerryOverlander I'm sorry to hear that you feel that way. I think it's important to remember that an index is not intended to be a transcription. The purpose of the index is so that the actual record can be found through a search and then viewed. There really are only a few items from any record that need to be indexed that people will try to search like, names, dates and places.
But the computer did index (or basically transcribe) the entire census. This hasn't been done before. Other censuses that have been indexed in the past only had specific parts indexed or it would have taken too long to make them available. And while the computer is not perfect, its accuracy is quite high. Which means that we can move faster and have a human review to catch what the computer didn't get right. But that wasn't enough for us. We have also created a process to allow corrections to the index after it's published. So there are multiple times to get this right. ๐คOh, and we didn't take a short cut on any of the information in the census. They will ALL be available. We have worked on some additional ways to get them reviewed (by humans) so we aren't losing any part of history. We are actually providing more in an index than for any other U.S. census.
I know you are worried about cursive writing. My kids can't read anything I write if I don't print - LOL. ๐But think about this. On this census, we are providing every part of the census in an index and as I said, that has never been done. There are many script types that have been used in history and I wasn't taught to read Gothic or other old script types, but I have learned and I think our kids will too. The desire to understand our family history is strong and I believe that will continue into our next generations and they will take the great things we have provided and do more.
4 -
@Sam Sulser @CaptBob Some good news! I was able to take the responses from everyone here further diving into reasoning for the changes and the explanations of how things will still be reviewed just differently and by others and provide my friends with them to take to their professors. I also took it to my librarian. They were so relieved! I believe they are reinstituting Familysearch as an approved source again!!! :D Thanks for all your continued followup on this for everyone, myself included. We all seem to also share the concern about cursive writing and out current and uture generations. I appreciate the stories about learning different types of writing such as Gothic and others as well :) I know we're gonna do a great job on this together.
One thing I think we can say together that we've learned on this journey is that more information in explanations of changes is best. The way it was described further in conversations here really cleared up so much for me and so many others. Especially the extended details about what will be happening and how this is the first such extensive AI reviewed and trained model ontop of the human interventions for it. Thank you again for all the time you've spent helping us all understand the changes better :) I give you all tons of credit!
4 -
@ Sam Sulser Thank you for your response and explanations. It has made me feel better about this process. I have question: Is there ever a way if we see a mistake in the future to change these? Or will it be once transcriptions are completed no changes can be made? I enjoy the process of transcribing and would love to help out any way I can.
0 -
That is great news! Thank you so much for following up with them and also with us to let us know.
@TerryOverlander Yes, the entire index will be editable in the near future! It is not at the moment, but they are working on the tool to allow this. Every version will be retained and searchable so that if things are changed for the worse, the prior version is searchable as well.
2 -
In Review Families, I have lately found that most of the families on the page are missing the surname, although it is clearly there in the census. I have not found a way to enter the surname where it says "Missing Field." I feel that I am wasting my time as I have to skip almost every family.
0 -
@Linda Mathew Please don't skip the families without a surname field. On the screen where the family names are usually corrected, click on "Report a Problem" at the top of that page and check the Missing Surname box to report it. Submit, then go ahead finish verifying the rest of the data. The surname still shows up in the census records. We are just reviewing data for accuracy.
2 -
The software does recognize that the family presented may continue on the next page and provides a mechanism for including lines from the following page.
The software does NOT recognize that the first family that appears on the page might have begun on the previous page. What do I do? Go ahead and identify the fragment as a family or skip that family and hope that whoever reviews the previous page will pick up the members that I skipped on the page that I reviewed?
1 -
One suggestion has been to click on Report A Problem (button at the top edge of the screen), click on Other, and explain it - I use language like the following:
The "household" identified by the AI is an incomplete household that is a continuation of the last household on the previous image/Census page.
1 -
@MichaelMayo4, when you are given a page where the first family on the page is a continuation from the previous family and has not yet been reviewed, you may report as explained above, but please go ahead and review the continued family. The system has been very good now at continuing to the next page to verify others who are members of that household.
Thank you for your concern and keep up the great work.
1 -
@AndLinda's comment/clarification is important because of the following:
When you "Report" the problem with the "household," you see a brief acknowledgment but then the system moves on and acts as if everything is OK. It can be confusing - it was to me. Did the system really understand what I just did? Why is it acting as if everything is normal with this household fragment?
But @AndLinda is telling us to complete reviewing/correcting the household "fragment" (and any other households we wish to on that census page). I take that to mean that the reviewed & corrected (by you) "fragment" eventually will (indeed) be reunited with its household mates because we've flagged it for special handling via "Report A Problem."
2