✨😀 IMPORTANT UPDATE 😀✨
Over the last few weeks, we’ve been reviewing your input about the experience and the work that has been reviewed so far and we’re excited to announce some major changes to the way things are being done!!! This update should happen tomorrow morning (May 24th).
First, Family Review is being simplified! There are quite a few fields that are over 98% accurate from the AI and you have verified that. So, we will no longer be reviewing those and some of the more challenging fields will be reviewed through new processes. YAH!! As such, the # of fields we are reviewing for Family Review is down to just 12! We hope that this will make it a much more enjoyable experience, one that we will be able to accomplish more quickly.
Second, Header Review, while still an important part of the entire review process, is going to be reviewed in a different manner. It will show as 100% complete, and you will be able to focus on either Name Review or Family Review.
Finally, Name Review will stay as is, we won’t see changes there—we are about at 55% complete! We anticipate being done in late June, and as we work to bring more people into this new experience, we hope to get it done even more quickly!
Rest assured, all the review done to this point is being saved and carried over into the other way that we will be reviewing these records. Nothing is being lost, your reviews have made it so that this next process will go that much faster. Again, thank you for all that you are doing! We’re excited to be able to begin sharing more frequent updates that show significant progress, especially in Family Review!
Comments
-
Fantastic improvements - especially for Family Review (to which I had recently suspended my 'personal' boycott by skipping most of that trivia - now being suspended!!!)
Thanks to FS do listening to users in this case!!
(Now if they would just listen to users about GEDCOM's. & the "new" & unimproved record search & ... ... Some folks always want MORE😞😊😊😊😞
1 -
I am puzzled by the statement that "quite a few fields are over 98 accurate from AI". That has not been my experience in the family reviews I have done. Addresses, state number codes, and particularly job descriptions. So many job descriptions "interpreted" by AI are totally laughable. Providing the AI interpretations of those fields to users is embarrassing.
7 -
Yes, many fields are not that accurate, and those will still be reviewed by another process. The fields that are are ones like the farm question, 3 acres, and some of the work questions.
3 -
I agree that it has not been my experience that the AI is over 98% accurate. That said, except for skipping the accuracy check of individuals' sex and marital status, I'm comfortable with the idea that the other fields don't have as high a priority. Others, like the state and occupation codes were entered later (not by the enumerators) based on the information previously entered, so they are just duplicate information that is not primary to the source.
2 -
The change happened so quickly. Just an hour ago I was checking all the fields. Then suddenly the fields were reduced. I thought something was seriously wrong with the program.
My only request is to add a notification on the Family Review page of the change so it's not a shock to others. A notification and link to this thread would reduce the confusion. Would that be possible?
BTW... I feel disappointed that I could no longer fix the job and industry fields. Frequently these fields were close but still incorrect.
8 -
I also agree as I have done work in many different aspects and it is definitely not 98% accurate. I have to correct nearly every document header in many ways and the family review does not get that information!!
5 -
The PRIMARY purpose of an INDEX is to lead users to the RECORD. THEY can then read all of the info, to their heart's content, and copy/record/use what they desire!!
I do, however, agree with the comment above by SvenThomasBerg that "individuals' sex and marital status" SHOULD be included - I've seen MANY marriages simply missed by AI!! (previously reported here & in Feedback)
This was a GREAT update & will greatly speed up the process AND lead USERS to the actual "SOURCE"
Addendum: it appears that the "Dwelling Serial Number" is always (so far) record as "1"
2 -
I agree that Family Review will go much faster by not including all those other fields that were extra, although many of those were not acccurate. I do agree with Sven that even if sex and marital status are 98% accurate, those are important for us to still check and be able to correct. Most of the time the AI is incorrect when the enumerator has crossed something out and rewritten it. Also the cursive 'mar' and 'nev' in the marriage column are often incorrectly interpreted by AI when the enumerator has poor handwriting.
And yes, this change should at least be posted in the updates feature, if not on the main page. I thought the system was acting up again.
3 -
I really miss the full field review, including occupations and income. It made it much more "real" and interesting for me as a reviewer and editor.
7 -
Hey All!!
I'm glad that you are liking the changes! We really do listen. We just don't always get to react this fast! You don't need to worry about the other fields. As I said some of them are coming in around 98% accurate and that was validated by you. Meaning they looked at what you have already reviewed and ran some statistics around what you changed and didn't change. But that wasn't true for all fields (as you've said above). There are some fields that will go through a different process to be completed. Trust me we worry about these things as much as you do and want to get this as perfect as we can!
We are so grateful for your efforts and your feedback!!
Sam 😃
5 -
Really disappointed you are taking away header review. I spent two hours today working on California. There are problems with AI reading census taker names and dates. Don't understand how this work will be completed if we aren't allowed to review the headers anymore.
Also we should be able to add missing surnames on both family review and name review. How do I know the error I report when a surname is missing will actually be corrected?
3 -
In my experience in reviewing 20000 names and 1300 families, the single most inaccurate field has been occupation. Not even close to 50% accurate. That being said, It's not that important
3 -
I'M NOT YELLING IN ALL CAPS, JUST LEGALLY BLIND.
I HAVE TO AGREE WITH MANY OTHER'S WHO HAVE COMMENTED ON THE ACCURACY OF THE AI INFORMATION. THE AI INFO IS TERRIBLY WRONG IN SO MANY WAYS. I'VE LOST TRACK OF HOW MANY TIMES I HAVE TO CORRECT SIMPLE SURNAMES MISINTERPERTED BY THE AI SYSTEM, EG., WILSON ON ON LINE, MILSEN ON THE NEXT, NILSON AND THEY ARE ALL PART OF THE SAME FAMILY WITH THE SAME SURNAME.
YOU HAVE ELIMINATED THE OCCUPATION AND INDUSTRY. THESE FIELDS HAVE BEEN INDEXED ON EVERY CENSUS. I REALIZE THE NEED TO GET THE CENSUS LAUNCHED FASTER BUT THIS IS A HORRIBLE ERROR.
DON'T GET ME WRONG I'M ALL FOR ELIMINATING ALL THE WORK CODES THEY WERE BEASTLY TO WORK WITH. BUT MANY PAGES OF THE CENSUS I'VE REVIEWED ONLY HAVE THE STATE CODE, IF THOSE ARE ELIMINATED, THAT INFORMATION WILL BE LOST TO THE END USER.
ADDITIONALLY, BY NOT INCLUDING THESE TWO KEY PARTS OF THE FAMILY REVIEW, THE NORMAL USER WILL NOT BE ABLE TO INTERPERT THE PENMANSHIP AS THIS HAS BEEN COMPLETELY ELIMINATED FROM OUR PUBLIC SCHOOLS FOR YEARS! IF WE DO NOT INCLUED THIS INFORMATION, IT IS ONLY DECIPHERABLE TO THOSE OF US WHO WERE TRAINED IN CURSIVE WRITING.
IN CONCLUSION, QUITE FRANKLY IT HAS TAKEN THE JOY OUT OF WORKING ON THE CENSUS AND ELIMINATED A VERY VALUABLE PIECES OF HISTORICAL INFORMATION. 😒
5 -
Thank you for listening to the comments. As one of the many who did not understand the overwork, this is great news!
This will greatly help keep me motivated to stick with the process.
Family review, as it was, was a tedious and overtaxing mess. Most of the omitted fields can still be reviewed by any researcher who needs them, by simply calling up the original image. (Doubt too many researchers will go into that much detail, but glad the image is there just in case.)
3 -
While I understand your reasoning, looking at the occupations, etc. made the job fun. You've taken some of the joy out of it. 😓
6 -
@RobertSteiner4 Just my perspective, but I have found the occupation field to be extremely helpful in some situations. When there has been ambiguity in associating a census record, I have found that the occupation field may be the clincher. I have wished that more people took that field seriously because of the help that it can be - as well as the insights into a person's life.
6 -
As noted by @Sam Sulser in the above comment, the occupation field will not be unindexed, "There are some fields that will go through a different process to be completed. Trust me we worry about these things as much as you do and want to get this as perfect as we can!"
0 -
@Teri Luna The header review will still be indexed. That is not being stopped. It will be completed in another way. Here is a short article about it.
Sam 😊
0 -
@KDKnight64 I think you may have misunderstood what is happening. We aren't eliminating any fields from the index. They will ALL be indexed. We only reduced the number of fields to review. Some of those fields were testing at 98% accuracy and will be published with the A.I. version. The fields that were not as accurate will be reviewed and corrected in a different way but will all be indexed. In addition, the ability to edit the fields once the censuses are published will still be available to catch anything that was missed. I hope this clears things up a bit for you.
Sam 🙂
1 -
I was disappointed to see the truncated family review process today. Why the rush to publish inaccurate data?
2 -
I have been working on one county at a time and it would help a lot to know the progress on a particular county. I can work days on a county and not see the overall progress for the state change. But I'm sure if I could see the status for that county, I would be seeing a change. And it would help with motivation to keep working on it.
This should happen when you select a county or city and it should give you current progress for what you just selected.
2 -
I hate this change so much. The AI is AWFUL! It never had anything right for any of the families I've reviewed beyond names. It gets it wrong about 98% of the time, not right. Plus all the information that is now not being indexed is just plain dumb. It makes it so that the information loses value completely for a lot of people. Not transcribing the full record makes it so you might as well not transcribe any of it. Big mistake. Big big mistake. Making it "easier" is not the answer here. If people don't like the process needed to do this, that's on them. It's important to have all of this information ACCURATE. Just hearing that this change is being made makes me doubt the validity of other information I've taken the word on that has been transcribed by Familysearch. It hurts your credibility majorly. I honest mean no disrespect with this and it's hard to tell over just reading it, but I have major concerns about this whole change here today.
0 -
Update, I informed my library about this change and they have said they will not be doing any more work or suggesting Familysearch any further unless this change is immediately rescinded or a further review is occurring. The AI is completely inaccurate and whoever gave you the 98% accurate testing is full of it. I'd say it's about 98% inaccurate to be honest and people may have just been skipping fields totally.
2 -
Most I know aren't liking the changes at all. In fact, it's hurting Family Search's reputation beyond this forum. As soon as my local library learned of this they removed Familysearch fromtheir recommendationss because they can't trust the accuracy of the information. They will change this if this change is rescinded but if not, I fear your information is only going to further decrease in quality. I love what we all do here together, but this was not the right approach. Adding a third step of further review may have been helpful, but completely eliminating human review of these fields which in the experience of most of the people here has been not 98% accurate or even close but far less. In fact, I know of peole who were just skipping the section entirely so it is inflating your numbers showing an accuracy rate that isn't in itself accurate. I've seen other reviewers do this quite a bit. I won't mention names here personally, but I'll tell you the number isn't small. It's certainly skewing your results in thinking it's 98% accurate. The harm to your reputation has already started from these changes and organizations outside are already taking notice at how alarming a change and statement that this was to make when individuals who work for these outside organizations can directly refute what you've stated. It is not good for the genealogical societies around, it's not good for anyone. I agree it was too much all at once to verify for alot of people, but that is why you make it another indexing opportunity as an addition to this part of it all. You don't just publish inaccurate AI information. In addition, a majority of younger generations can't even read the cursive writing in the documents to begin with. You say it isn't being eliminated but it might as well be because kids don't learn cursive anymore. They physically can't verify what you're showing them to be true. That is a huge problem. If the census were all in standard writing this would be one thing, but to do this with a document that has cursive in it isinexcusablee in 2022. I am not saying this to be mean spirited or argumentative or to break any rules. I am genuinely concerned for the organization's credibility from this because I'm already seeing the effects of this change and it hasn't even been published yet.
3 -
It's interesting to see the comparison between YOUR comments & analysis of the Family Review changes with others posting to this thread - a fairly large difference!!
I found it a bit strange that you apparently "informed (your) library about this change" BEFORE you posted your comments to this Forum. You must have a great influence on your library if, based solely(?) on your comments (complaints?), "they have said they will not be doing any more work or suggesting Familysearch."
BUT, I'm glad to hear that you are "not saying this to be mean spirited or argumentative" & that you are "genuinely concerned for the organization's credibility." (The "organization" is, of course, called FamilySearch, operating as an entity created by The Church of Jesus Christ of Latter Day Saints.) I might suggest your actions do not seem consistent with your comment, but I could be misunderstanding your rationale for so vehemently disliking the change that has been made.
I gather you will not, or can not, name what "organizations outside are already taking notice at how alarming a change and statement that this was to make" - I'll freely admit that I think it was a "Fantastic improvements - especially for Family Review" as you can see from the first comment posted above! I do have the feeling that we may DISAGREE!
It's also interesting that apparently you just joined this forum today simply to post this opinion - perhaps you should read a bit more in the forum, it sounds like you are reading a lot elsewhere!
0 -
I like the simplfied family review. I do not know why this causes concern with some. It will all be reviewed.
1 -
The data is there. What review that has been done to this point is not lost or wasted.
The images are there. In any case, when the indexes are published, we can go back to the original images to clarify any issues.
We now do not need to deal with all the fields. (Especially the mind-numbing minutia.) Others will.
The review process should complete quicker.
If one reads the **entire content** of the message, the people running this show are trying to help us!
There are other processes being applied to this project; they are respecting our time and contribution.
For some, it is hard to accept any change - even if it is in our favor.
Thank you to our project teams as you are trying to help us and get though the data quicker and more accurately by applying other resources to this task.
2 -
Hello there. I was working on this in my library as was a librarian at the time. They also contribute regularly. We were sitting together when I asked her if she was having any issues trying to correct parts of the family review process. She actually informed me. I may have misworded that above. And she said that she only just discovered this and showed me the community forum post. She was waiting to hear back from the Genealogy department's head to determine what course of action is best to take. That's when she got a message from them that they were enacting this to remove FamilySearch for the time being until they can verify themselves whether or not this AI model is TRULY 98% accurate in these fields. They highly doubt that is the case. I honestly thought I had a community account as well, but apparently didn't. I never really check into the community that often, but at times I do. Always thought it was signed in but apparently not. It's a bigger issue than you're realizing. People are noticing outside the community. I didn't even know about it yet when I first came across it and was suggested to check the community pages.
I personally do love FamilySearch and I honestly believe this was made with the right intentions, but intentions aren't always the same as the results. Not even 24 hours later, a friend of mine who also contributes was contacted by a professor of theirs and a new policy instituted in the Sociology and Anthropology department at his university not to use FamilySearch.org records for the time being until they can verify the accuracy of the AI model. I have since sent the message that was sent to me from them to FamilySearch so that they can do with it what they will.
I want to be perfectly clear that while I am unhappy with the change, I do see the merits of SOME of it. But do believe a third workflow should then be added to make sure these fields have human review by those willing to go through them individually. One concern brought up by my friend's professor was also that students coming into their classes now no longer can read cursive writing. Much of the census is done in cursive. They don't want their students using resources that they can't personally verify to be valid in regards to cross checking what is transcribed versus what the document itself seems to say. There's a lot more that goes into this than meets the eye is my point overall.
I appreciate the dialogue with you. It's hard because the way we type isn't necessarily the way we speak, nor does it carry our tones all the time so we're limited in expressing things. Like I said I'm not trying to be mean spirited or anything at all. I'm just seeing effects of this and wanted to say something so somebody knows what is going on here and what our changes may impact that was unintended. :) I value our work as you do and know that regardless of changes we will make it as accurate as we are allowed to make it. But that's the issue. Allowed to make it. Having the ability to correct parts of the forms that are incorrect was and is important especially in occupational fields when it comes to anthropological and sociological study. It also helps verify a person as the accurate person in a family tree in many instances. Inaccurate and sloppy AI information discredits the work we all do together. AI is not supposed to be just left to its own accord. It must be properly monitored.
In fact, as even evidenced in some comments above people were skipping these fields entirely thus inflating the AI model success rate since it was being determined based off of what fields were actually changed and which had no changes. If you've got people skipping fields it's then going to look like you've got the greatest AI model of all time with that being your way of saying it's a great model. When in reality you could have the worst AI model around. Human verification is paramount in ANY historical document transcription.
1 -
Thank you for the prompt response and clarification, Ian. My apologies for my possibly 'harsh' comments to your post.
To me the key to understanding the BENEFITS of this indexing changes is the efficiency that is no being offered to us, the indexers. It will decrease the time expended quite significantly!!!
IMO, the PRIMARY purpose of an index is to direct a researcher TO the source, which can then be thoroughly analyzed whether the AI was 98% or 42%. NO information will be lost from the census page (just less items indexed by a person). The user/researcher can then decide the reliability & usability of the information and use it as they see fit.
The "human verification" will then be done by the PERSON who is truly interested in the record as that indexed person in the record is (likely) their relative, or person in whom they were interested; not simply a person indexing a record to assist others in finding said records. An INDEX is a finding tool and does not need to be (and should NOT be!) a transcript. The 1950 Census will stand as a record in, & of itself, regardless of who (or what) provided the index!
I will now return to the expedited indexing now available for 'Family Review' that I had been personally boycotting due to the complexity and time required! 😊
Now that you are a community forum member, I hope you will continue to be active herein.
6