Home› Welcome to the FamilySearch Community!› Ask a Question› Get Involved/Indexing

Review Names, read outside the box?

dontiknowyou
dontiknowyou ✭✭✭✭✭
March 21, 2022 edited April 10 in Get Involved/Indexing

I am seeing names where the white box contains more than the blue box. I think this is due to other contributors reading outside the box. Reading outside the box probably is a mistake, since the AI is being trained on what is inside the box.

Here is an example. I am going to correct this to "Hess".

Screen Shot 2022-03-21 at 3.36.32 PM.png


0

Answers

  • dontiknowyou
    dontiknowyou ✭✭✭✭✭
    March 21, 2022 edited March 21, 2022

    Similarly, here only "Hess" is in the box but the suggestion is "S. C. Hess". Correct that to "Hess". But, I am going to flag this one a transcription error because the box is reading also the line below.

    Screen Shot 2022-03-21 at 3.41.32 PM.png


    0
  • ctr sweetie
    ctr sweetie ✭✭✭
    March 24, 2022 edited March 28, 2022

    @dontiknowyou , it appears that you corrected them appropriately in both instances. If the highlight covered part of Elizabeth, too, you could include her first and last names, Elizabeth Hess. In the second example, marking a transcription error is also appropriate for the reason you indicated. Thanks for including the screenshots in your question!

    0
  • JoanneMacD
    JoanneMacD ✭
    March 24, 2022
    https://community.familysearch.org/en/discussion/comment/438361#Comment_438361

    I'm a bit confused by this. The OP said that she took out the first name or the initials because they were not in the box. Then you said that she "corrected them appropriately" because "correcting it to the complete name is appropriate". From what I can tell, what she did is the opposite of what you said to do. Which is correct?

    Thanks. I'm loving this new option for when I'm on my phone, but I want to make sure that I do it and teach others properly.

    0
  • dontiknowyou
    dontiknowyou ✭✭✭✭✭
    March 24, 2022

    Keeping in mind that we are training an AI. It seems to me that we should teach it to correctly read what is in the box and only what is in the box.

    I think correcting "Hess" to "Elizabeth Hess" would be an error. If I train the AI to read "Hess" as "Elizabeth Hess", what happens when the AI encounters the "Hess" part of, say, "Heinrich von Hess"?

    0
  • Elmer
    Elmer ✭
    March 24, 2022 edited March 24, 2022

    I think you are totally misreading this.

    The AI is not looking for what is in the blue box. The AI is looking for names in the document. The AI finds a name and then highlights it blue.

    So, the question is, was the AI looking for a full name and completely miss the given name and only highlight the surname?

    Or, was the AI looking for just individual given and surnames? Did it already highlight the given name previously, and then correctly highlight the surname separately?

    That is the information we need to judge correctly. Otherwise the reviewing of records is really going to be messed up with everyone interpreting things differently.

    0
  • dontiknowyou
    dontiknowyou ✭✭✭✭✭
    March 25, 2022

    I think you are totally misreading this.

    Who? Misreading what?

    Moving on...

    The AI frequently picks out only parts of names. When I submit the text in the blue box, and only the text in the blue box, then the AI assembles boxes into the complete name. This is how we can train the AI to assemble complete names.

    Validating names the AI fails to detect merely trains the AI to make errors.

    0
This discussion has been closed.
Clear
No Groups Found

Categories

  • All Categories
  • 42.7K Ask a Question
  • 3.3K General Questions
  • 570 FamilySearch Center
  • 6.7K Get Involved/Indexing
  • 640 FamilySearch Account
  • 6.5K Family Tree
  • 5.1K Search
  • 996 Memories
  • 2 Suggest an Idea
  • 473 Other Languages
  • 62 Community News
  • Groups