I stopped by MH to do my periodic testing of their product.
Found that Record Matching had actually produced some useful results, for example an 1880 census for my grandmother. I had found this 40 years ago in physical search, and a couple years ago in other online searches, but it is good to see that MH is catching up.
So I copied the data into her record....only to find that the code had not ADDED new data but had REPLACED data from the 1906 Prairie census of Canada.
Unfortunately, that was the only record available so I couldn't try another one to see if I could replicate the error.
So, user beware. MH seems to allow you only ONE record per individual.
This is an egregious error that needs to be fixed as soon as possible.
The extract page tries to match information from the record with information on the profile. In cases where information already exists on the profile and the user wants to add the information on the record as an addition to what already exists on the profile, it can easily be done by clicking on "Show more options" and checking the "Add as alternate" checkbox. This will extract the new information in addition to what already exists and will not override the information on the individual's profile.
I urge your developers to set "Add" as the default option so that the user does not have to take two more actions, neither particularly obvious, to add new information.
I would argue that new data is not "alternate".
The 1880 census is not an alternate to the 1870 census. It's entirely new information.
If I already have a date and place for a marriage from the wedding certificate, and I find a different date in a newspaper announcement, that is alternate information.
If the information in the newspaper is identical to that of the certificate, that is duplicate information but it is equally important because it is additional evidence to the accuracy of the information.
If it is duplicate info, it should not be added but only the source cited as additional support for the data. Don't double up the data. Ancestry does that and it's annoying because I have to move the source citation over to the original data and delete the duplicate (If I don't, it clutters up reports no end). Here's a chance for you to beat the competition!
I cannot think why you would have done it the way you did but I know you're working to improve so here are further suggestions.
Got a media release today that said, in part, We're excited to announce that in the coming days and weeks more than 2 billion additional records will become available to search on MyHeritage. This is the result of a new partnership with the leading genealogy organization, FamilySearch....
Our matching technologies will be unleashed on these new records and profiles so MyHeritage users will receive Smart Matches™ and Record Matches whenever any of them match their family trees.
Wow, two billion more records! Whooee!
My fear is that unless the new records are accompanied by improved search engines for SuperSearch, SmartMatch, and Record Detective, the result will be two billion more potentially irrelevant matches to clutter things up.
Hey, maybe they will include US state censuses and Canadian records, both sorely lacking at MH.
When I use the research function it brings up possible death certificate matches but NO additional informaion like parent name, spouse ect. Unless I order and pay for dozens of potential matches certificates I still don't know if they are my ancestors. Am I missing something?
When I use the research function it brings up possible death certificate matches but NO additional informaion like parent name, spouse ect.
By "research function" I'll assume you mean SuperSearch. You might have to be more specific. How did you get to "possible death certificate matches"?
When I try this, by drilling down All Collections > Birth, Marriage, Death > Death, Burial, Cemeteries, and Obituaries > (for example, California Death Index) I find date of birth, date of death, place of last residence; on accessing the record I also find name of father, name of mother. Oh, and I spotted one that says "Missouri Death Certificates" showing date and place of death. When I click it, It shows me the actual record. But then, I have a data subscription.
So I guess the answer to your second question is, Yes, unless you order and pay for data access, you won't get the additional information to determine if you have a match.
I am now starting to get UK census data record matches in my tree, which I can then add to my tree. Some of the names are in both the 1871 and 1841 data but if I try to add the second census data it seems to want to overwrite the original data rather than adding to it. I may have missed something here but can this be done from the records info rather than having to do it manually.
The record matches are matches with the details in the specific persons card in your family tree.There are not many cases where the persons card matches the results of both the 1871 and 1841UK censuses.If you want to add the 1871 census data to this persons card then your approach of
manually adding this information is correct as it can not be done from the records.
I'm having exactly this same issue with the US Census data. And since those are only 10 years apart, I have a great many people in my tree with numerous US Census record matches. This is a big issue, I hope MyHeritage is working hard to fix it. Manual entry is just much more difficult when you consider everything that the Record Matching automatically does.
As there is only 30 years between the census dates I suspect there may be quite a few instances of the same person appearing in both sets of data. Occording to the info on the Research site the plan is to make more of the UK census data available in the linked data area so this is likely to occur more often in the future. It would seem a sensible inhancement to the system to allow more than one year census data to be added as the citation source can be selected in FTB.
The new SuperSearch ability to attach a document/record to one or more individuals is a step in the right direction. The ability to extract information into a record has been iffy in my tests, but will be great once it's ironed out.
The issues I incountered may due in part to my inability to find much of value with Supersearch. In one test, I got a message saying "Extract Records is coming soon" and in another case the extract proceded, only to later give an error messaging saying it was unable to save the source information (though when I checked, it appeared to have saved). Both these issues were reported to Support.
A recent supersearch turned up 70,000 newspaper clips. You'd think some might be of interest, but...
They're so disorganized! I have no idea what order they're in. They're not alphabetical by name. There needs to be some way to sort those newspaper reports
by date to focus on those within my target's life span (that is, a date range rather than simply a date of publication)
geographically to concentrate on those where he is known to have lived (I know, I can repeat the search with each known location, but a sort would be nice).
I did skim the first 60 and found a couple that looked possibly relevant for my target (J.K.Gray) But I found that in the first, the J was in one place, the K in another, and the Gray further down the page. Not too useful. The other was 29 years after my target's death (can't eliminate it just for that) but turned somebody I knew of down in Calgary (a namesake but not a relative) so I didn't even open it. Ah, well, another 30 minutes wasted.
They only have one newspaper from the province of Alberta, and it's from a smaller center several hundred miles from where my family lived, so I'll just have to keep trying in case they get some records from other cities. Another time, I'll repeat the search state by state to see if that helps too.
Thank you for writing us some more suggestions about the newspaper archive. I'll be sure to write them in the long comprehensive list of suggestions that you have already sent us. Our developers will go through each of the points.
The sorting of the newspaper archive is being done by the most relevance for the search parameters that you have entered recognized by our system.
Since the newspapers are unstructured records, it is harder for the system to recognize the grouping of letters and so problems such as the one that you have described are present. We are working to improve the algorithm further though it may take awhile.
Wow, 2.56 million results in 1.31 seconds! That would be impressive if they were relevant.
Well, okay, the first six are relevant, but they're all family trees, most of them managed by my relatives. And a hit from Find A Grave, which I found some years ago all on my own. And no way to eliminate them from future searches.
As for the rest...
Millions of people with the wrong name, the wrong birth date, the wrong death date. WHY?
Hundreds of thousands of census records from after my target's death and before his birth. WHY?.
Birth and Death records from all over the world, even though I've specified the places for each. WHY?
World War II service records for a man who died in 1937. WHY? (this strikes me as particularly stupid!)
Students from Queensland, Australia for a man b. Kansas, d. Alberta, never left the continent. WHY?
Stupidsearch. Ignore the search parameters, throw in everything no matter how irrelevant.
I'd rather have 100 likely candidates instead of millions of pieces of junk. I wish I could make this thing actually heed the parameters I put in, instead of ignoring them (or, if I tell it to behave itself and use exactly what I enter, give me no results at all). Gotta be something in between no results and 2.56 million.
I'm sure that a lot of members are confused as to why they would get so many unrelevant results for the search parameters they input. MyHeritage's SuperSearch is not that different from Google's, Bing's or any other standard search engine in the sense of functionality. The search engine will try to find all related information that could fit some or all parameters you have entered.
The search engine will also rate the results with internal logic and will show you the most relevant records on the top, however as you scroll lower you will see less and less relevancy to the search that you have inputted.
Unlike other search engines we provide with a lot of different fields to try and search for the records a little bit further. However, if you input these fields it doesn't mean that they are necessarily present in all of the records. For example a UK census record may not have information about the spouses names. Our search engine tries to get around this by including the census record and to ignore the parameter that you have inputted regarding spouses. Thus showing a lot more records.
You have mentioned that you have problems with the "Place" field, again, as some records may not have this information and immigration results will logically have to be shown (in case this person may have immigrated to other countries).
There are many other different reasons for receiving matches that seem completely unrelated:
Names - We hold a big dictionary of name variances, and when you search we will try to provide you with all the records including the name variances of this person.
Dates - Our supersearch searches for dates that are around the one that you specified ( /- several years) just to make sure you don't miss out on these records as well!
Since we show so many results we have two work-arounds to try and effectively go around this problem:
1) Post filtering - you can use the buttons on the bottom left of the screen of the SuperSearch to add an extra filtering to the records (for example filter all the results for the first name of "John" to be matched exactly).
2) Specifying the search (on the main search page) as "Exact". You may have noticed that we have added this option recently. There's a little checkbox that allows you to search for the parameters you have specified exactly. The system will not ignore these parameters if you specify the search as an exact search, meaning that you will get a lot less records.
If after searching in the "Exact" search, using the post filtering to search in the non-exact search and going through some of the first pages of the search you will not be able to find anything then you will have to consider the possibility that we may still do not have the records that you are looking for.
Then you will most likely have to wait until we acquire these records (which we continuously add) or check the smart/record matches.
P.S. Could you perhaps provide me with the record in point #4 to our support email so I could inspect if further?
P.S. Could you perhaps provide me with the record in point #4 to our support email so I could inspect if further?
By the way, using Exact as you state almost invariably gives no results. It's entirely possible that for any precisely specified search there are no valid results, and that's fine.
Even given your explanation above, though, I can't see why the code should report
149,000 results from the 1940 census for a man who died in 1937. There is no way that I can conceive of any of those being even remotely relevant.
32,000 results in the 1850 census for a man born in 1856. He wasn't even conceived then.
22,000 results for the 1871 United Kingdom census for a man born in Indiana, USA. I don't expect to find anything useful there. If I hadn't specified the place of birth, I could see it. But the search does know where he is born, so why include this at all?
I'm just continually confused that these sorts of things are included in a so-called smart search. I think it's a kitchen-sink approach ("throw in everything including the kitchen sink")
I think it would be a great idea to give us an additional choice of "possible" so that we can see which ones we have already looked at but were unable to rule out or confirm. With the number of matches with <private> names, many matches can't be confirmed or ruled out, yet they stay on the list without any note that they have been looked at. Can you make it possible to mark the ones already looked at?
What I do, in the cases you described, is go to compare, then scroll down to the bottom of the page and leave a message that this could be a match, but unable to confirm due to "private", unverifiable entries, then back arrow and reject the match. Unless the other site updates and removes the "private" situation, it is useless to keep the match on my site. Sometimes, the other site manager contacts me to give up the "private" info, at which point I find the rejected match and change it to accepted. Charles.