Log in
Sign up
View all  |  Close all
#
Subject
Posted by
Actions
1
Where is the LOGIC in Smart Matches
BrianB Andrews
- Close
While I see the benefits in smart matcing especially in connecting to researchers with common interests and also being able to share information.

What I cannot follow is the logic or process used for recognising a Smart Match and posting it; I have spent almost an hour this morning going through the procces of ignoring other trees nearly all Scandinavian that showed matches for a Lars Larsen in my tree and what appeared to just be anybody from their tree there was no Lars or Larsen in the supposed to be match name, B & D dates were not matched and other information bore no relation to the information for my Lars Larsen! I thought it was never going to end and the list seemed to get longer as I ignored each tree.

After many years in the IT business it is quite difficult to fathom the basis and logic MH programmers use to determine a SM are they trying to give us all RSI!!! We could turn SM off but then we miss out on the genuine matches and connections out there, what happened this morning is quite a frequent occurence especially with FT's of Scandinavian origin and no doubt they must be suffering the same as well.

MH certainly needs to lift its game and give its members better and more ifficient results and service in this area at least give us more options to filter the SM i.e. reject anything less than 70% match, I certainly find the Ancestry HINT's more managable; the 2 step priocess in MH is rediculous and onerous.

     

BrianB Andrews
New Zealand
- Reply
2
RE: Where is the LOGIC in Smart Matches
Eric Dalton
- Close
SM does not do Scandanavian names well at all. I made a comment on this a long time ago, when someone was talking about just this sort of thing.
Eric Dalton
- Reply
3
RE: Where is the LOGIC in Smart Matches
qbuster
- Close

I agree and others should be aware that this has been an issue raised in forums for at least 2 years.

To be fair it is measurably better than it was but as you point out as another (now retired) programmer I just cannot see why Smart Matches ignore the obvious - surely a profile from tree that has 99% percent of its relatives in, say the UK, is matched with a surname with obviously different spelling from another country should not be matched with such a high rating this particularly when the B&D aren't even within 10 years.

qbuster
Alrewas, Staffordshire
- Reply
4
RE: Where is the LOGIC in Smart Matches [2]
Daniel D.
- Close

Hi,

We are working on improving the Smart Matches system to be as accurate as possible. We have recently made big improvements to the sytem that are yet to be updated. We are still testing the improvements we have made for the system and it should come out soon. Our accuracy rates (from what we can see) have reached 97%. We have tested it against trees that had a lot of rejected matches.

I would like to let you know that Birth and Death dates are not necessarily known to the individual who researches the family especially if they are not closely related to each other. Names can be spelled differently by different cultures and by different people. We are trying to get individuals with as much revealing smart matches as possible rather than matching the same people with the same information.

Along with our updates to the new system not only the matches will be more high quality, but also the diversity and the information you can gain from them should be enough to help your research. 

We are doing our best to supply you with the best technologies possible. And we are interested in promoting genealogy to everyone with our tools. 

Daniel D.
- Reply
5
RE: Where is the LOGIC in Smart Matches [3]
BrianB Andrews
- Close
I do not know what cue card you are quoting from Daniel; from the coal face i.e. being a daily user of MH for my family research with less than 2 years experience I have seen a gradual decline in the quality and standard of the matches being thrown up for my FT since those early days. It has got much worse over recent times that I have felt fed up and frustrated enough to express my concerns by starting this thread.This mornings trio of matches bear no matching qualities at all relating to the name and I dispute your comment relating to the name being spelt differntly with countries and cultures it is the born name that is important no matter what country of culture i.e. a person with a given name of Annie would be Annie in NZ & in Sweden and in reverse a Gustafa born in Sweden and connected or in NZ would stll be Gustafa. The examples below should not be matches. Annie Palmer born Mason 1888-1967 v Maria Elvira Johansson 1884-1884 Annie Palmer born Mason 1888-1967 v Hulda Maria Carlsdotter 1884 – 1884Annie Palmer born Mason 1888-1967 v Gustafa Tvilling med Maria Carlsdotter 1884 – 1884 You or MH may feel they you are making improvements, Hello, it is not happening MH need to engage and communicate with their user base about these so called improvements all I am seeing is more down time on my part wading through what appears to be illogical mismatches rather than focusing on my primary goal to grow my family tree. It is not all bad as among the matches there are genuine ones the mismatches are getting worse more frequent especially Scandinavian to English
BrianB Andrews
New Zealand
- Reply
6
RE: Where is the LOGIC in Smart Matches [4]
Daniel D.
- Close

Hi Brian,

We actually worked on smart matches such as these to check whether or not our smart matches system gives better accuracy and results after our improvements. And it does.

Looking into your specific situation I would be glad if you could give me links to the smart matches to see if I can find any reasons as why you would get these matches. I tried doing some investigating already, but I couldn't find Annie Palmer's bad matches in your smart matches - all I could find are 4 confirmed ones.

It is important to note that we did not yet update the Smart Matches in our major update. We are working to perfect it and the system before we release it. Also we are working on the annoyances such as where you would go to the smart matching page and you would see no pending matches (even though the counter says there are new ones) - we are on the verge of resolving this (and it will be present as soon as possible).

Daniel D.
- Reply
7
RE: Where is the LOGIC in Smart Matches [5]
BrianB Andrews
- Close
Hi and today there is another 66 matches recorded in my tree the majority of these are garbage and as I have already pointed it is very difficult to follow what logic is being used to generate these matches. The majority of these are scandinavian names matched to english and as i have already stated a person given a english name at birth retains that name no mater where they live you have no concept of how frustrating this is and the effort it takes to physically select each of the matches individually and then ignore the other tree which is a 2 set process.

Is there no testing done before the matches are put into production IU cannot believe that this is just happening to me alone.

There have been fewer rediculous matches these past few weeks probably one or 2 a day but to get a bunch of 66 in a day, something is going wrong!!!

 

 

BrianB Andrews
New Zealand
- Reply
8
RE: Where is the LOGIC in Smart Matches [5]
BrianB Andrews
- Close

Hi Daniel there is another example of those mismatch examples I gave 2 days ago on my FT today I will leave it there as an example for you to see i.e. i will not do anything with it

it is

Annie Palmer born Mason in my tree

Gustafa Tvilling med Maria Carlsdotter in Familjen Samuel Karlsson Site 

Cheers Brian 

BrianB Andrews
New Zealand
- Reply
9
RE: Where is the LOGIC in Smart Matches [5]
BrianB Andrews
- Close
Hi Daniel thanks for your reply I actually got rid of those erroneous matches by selecting the 'ignore this tree' option a few hours ago I am not sure if you can track this via your logs! below is an example I located; I only rejected this several months back “Jessie Christina Mayhill (born Ross)” in My Treewas matched to Alpina Emma Settimini (geb. Rossi) In Family site: Familienseite Katrin Thiel's from Germany's tree I only rejected this match so it is still in my list.This is a little different as the DOB & DOD years are the same for these two people however I still expected the SM criteria and logic to first check the family and given names and reject at that point . Cheers Brian
BrianB Andrews
New Zealand
- Reply
10
RE: Where is the LOGIC in Smart Matches [3]
fps
- Close

Hi,

this sounds good - let's see how it works.

Anything new should be an improvement over the present situation giving matches even for profiles of different gender wirh almost no similarity in the names.

Also needed is improved privacy - hopefully, this will be included in the new system. What sense does it make that someone can copy, via FTB Smart Matching, a complete family including all living persons? After this happened to me, I stopped using Smart Matching.

 

 

fps
Lümborch
- Reply
 
Loading...
Loading...