Grenn temple icons and duplicates

LegacyUser

Glenn P said: Here is an example of a duplicate situation, where no record hints (Research Help) have come up to show the duplicate, either for the husband or the wife on their person page.
The original PID dated LCJD-4FK Michael Dafoe, UE. (UE is a designation for ‘United Empire Loyalists’) PID created April 2012 by FamilySearch. Lots of action on the PID since. Temple ordinances completed 1978 – 1997. Note no record hints (Pending Helps) for Duplicates.
The spouse to this PID is Rachel Holcombe-Ross LZ6V-YN6. Temple ordinances completed 1973 – 1991. Note no record hints (Pending Helps) for Duplicates.
And now the duplicates created June 23, 2020:
Michael Dafoe PID GWMX-VH4. Ordinances reserved June 23, 2020, shared with the temple, icons all Green ‘ready’ except SP.
Rachel Holcombe-Ross GWMX-8XF. Ordinances reserved June 23, 2020, shared with the temple, icons all Green ‘ready’.
As it now stands I can pull the reserved temple ordinances for the second Michael and Rachel, and when the temples are open, print and take to the temple. But I will be duplicating all ordinances for two people. (I am also aware of the 90 day limitations).
Of note the person who created Michael Dafoe GWMX-VH4 and Rachel Holcombe-Ross GWMX-8XF on June 23, 2020 (and reserved them and shared with the temple), also added Sources to the original PID LCJD-4FK on the same date.
When I use ‘Find Similar People’ from the person page for Michael Dafoe GWMX-VH4, Micheal Dafoe ‘Sr’ LCJD-4FK and spouse Rachel Holcombe-Ross LZ6V-YN6 show up. (For some reason not clear to me the ‘UE’ is now showing as ‘Sr’ in the people found list [one of his alternate names], yet opens to the person page as Michael Dafoe,UE).
My concern is that there is now temple work ready-to-go Green for the second Husband and wife. I realize also by checking on ‘edit name’ that the UE was included as a part of the last name rather than being added as a suffix. However, would this also affect the system being able to find the 2 Rachel Holcombe-Ross PIDs as duplicates? Perhaps it has to do with standardized locations. Yet when I use the ‘Find similar people’ function, they show up.
I am wondering if anyone can help me out to understand why the duplicates did not show up as possible duplicates in ‘Research Help’ on the person page for any one of the 4 PIDs and yet show up using the ‘Find Similar People’ function? I also realize these are being found as ‘similar people’ and not necessarily as potential duplicates. Perhaps the search parameters for duplicates are more limited. My understanding is that if the possible duplicates did show up in the ‘Research Help’ hints, the temple icons would not have gone Green (unless someone actively dismissed the ‘possible duplicates’ hint). Yes, I can clean it up and contact the other patron. I am just wondering about the present situation and 'Research Help' for duplicates.

LegacyUser

gasmodels said: I am no expert on this issue but do know that possible duplicates are determined by a scoring system where if the score is sufficiently high the two records are considered possible duplicates. Points are based on a variety of pieces of information being the same. Lets take the spouses as an example and do a quick comparison and you may see why they do not show as possible duplicates. This is a very sensitive issue because it has been shown that many inexperienced users will merge every possible duplicate listed without any concern for the word "possible", therefore the matching tool is somewhat restrictive and the "find similar people" function was added for the more experienced users to look further for potential duplicate records.
Comparing Rachel Holcombe-Ross LZ6V-YN6 and Rachel Holcombe-Ross GWMX-8XF we can see the names are the same and the birth and death years are the same. However, one is born in New Hampshire and the other in Vermont. The spouses have the same name except one has a ,UE attached to the name which the system might consider a different spouse. Entering the UE as a suffix rather that part of the surname would probably increase the spouse scoring. LZ6V-YN6 has a mother attached whereas GWMX-8XF does not. There is a single child associated with GWMX-8XF which does not match any of the children associated with LZ6V-YN6. The death location on one is not standardized which may cause the duplicate feature to not do a proper comparison. I think you can see there are probably enough differences so that the comparison algorithm has a score too low to indicate these are possible duplicates. It would be interesting to see how many adjustments would be required before the two records displayed as possible duplicates. It could be that only minimal adjustments would be required.

I believe a similar analysis could be made for the spouse. Like I stated I do not all the details but it does not seem unreasonable to me that these two records do not appear as possible duplicates.

LegacyUser

Tom Huber said: Whenever I review a record, I will use the feature "Find Similar People" to have the system search the massive tree.

That search uses different criteria that the possible duplicates routines, which of necessity, has to be restricted.

Far too many users see a possible duplicate and never consider that the "duplicate" may not be a duplicate record.

LegacyUser

Gordon Collett said: I also have no idea what the criteria is for declaring two records as possible duplicates, but if you spend much time on this board, you will see why it needs to be very strict.

A common complaint here concerns incorrect merges that were performed based on the reasoning that the computer said it was a duplicate. Users here have been very frustrated at times because other users look at "Possible Duplicate," ignore the "Possible" part, and merge without further thought. Some of the features of the new merge routine have been put in place to try to prevent this.

The goal of no bad merges would lead to increasing strictness of the possible duplicates routine and lead to more missed possible duplicates.

The goal of no missed possible duplicates would lead to decreasing strictness of the possible duplicates routine and lead to more incorrect suggestions for duplicates and increase potential for more bad merges.

To set the limits for the possible duplicates routine one must decide which is worse: missing some duplicates which can be found in other ways and risking repeating some ordinance work unnecessarily or showing all possible duplicates along with a lot of false duplicates and having a user's careful research merged away leading to someone's ordinance work never being done?

The current system allows for three levels of research experience:

1) Little to no experience - Possible Duplicates routine should provide, ideally, only high quality results that are truly duplicates as close to all of the time as possible. (It hasn't achieved this yet.)

2) Moderate experience - Find Similar People provides a large number of matches that superficially look the same and need to be carefully evaluated to see if duplicates or not.

3) High experience - Find allows skilled researchers access to the entire database to find even the most unlikely looking true duplicates

LegacyUser

Glenn P said: Thank you gasmodels, Tom and Gordon for your comments and observations. Much appreciated.