Use AI for learning for manual correction and links
In a lot of cases the indexing is terribly wrong because of the missing knowledge of the local languages and not understanding the manual writings, additional notes, comments, since it would require the language knowledge. The local language would help to distinguish between correct and mistyped family or first names. You also have to know the culture of time, how the writing standards changed. An indexer should also know the local Latin dialect, not just the native languages.
However, if you could teach an AI system to use the manual correct links into the manually corrected and maintained records, then you could automate a lot of correcting tasks for the indexes.
Comments
-
Yeah, no. FamilySearch has an abysmal record with automated anything; witness especially the autostandardization flustercluck that has rendered location fields in the entire indexed records database completely useless. (Yes, some of them are mostly OK, but some of them are only on the correct planet because that's all that's in the database, and when doing a search, there's no way to tell which is which.)
1 -
Wish it was possible to record a downvote. As Julia said, we're already dealing with a lot of mess from automated standardization. No need to add more.
1 -
Testing on beta system would be ok - but not production environment until perfected?
0
