Fra Noi Embrace Your Inner Italian
Big news! Some of it great, some of it…not so good.
When I was at the Rootstech conference in Salt Lake City in 2024, Familysearch was making a big deal about using “AI” technology to transcribe genealogy documents into searchable data.
They gave us a taste by indexing words from wills and probate records which are typed on a typewriter rather than handwritten. My genealogy colleagues who need wills and probate records were all excited, and rightly so. My Italian genealogy colleagues were expectedly disappointed.
I’ll try to explain this for people who are not tech nerds like me. The original documents housed in libraries and government offices were photographed and posted on FamilySearch. But a human being would have to read each record and type into a computer the name, date, location, parents, and other data from each photograph. Simply put, you can’t search a photograph of a piece of paper for the words that are in it.
The process of human beings looking at each photograph one by one will take forever. There are uncountable billions of images, and the humans who review them are volunteers. Even though a lot of people are willing to help, most of us don’t have time to give for free.
That said, if they can teach a computer to “read” each document and extract the names, dates etc. from it, it would be much faster and they could put as many computers to work as they like. Computers don’t need coffee breaks or vacation or sick time! They work around the clock and they do not get tired like me!
Sounds great, but how does this work in reality? The computer has to look at each letter of each word on each document, and determine what that letter is. It could be a typewriter letter “a” or it could be handwritten “a.” Typewriter script is usually consistent, so the “a” on one document looks mostly like the “a” on other documents. When the computer examines the shape of the letter “a” it checks that shape against what an “a” should look like, and reports that the letter is an “a.” It does this to the next letter and on and on until it finishes one page, then it goes to the next page. Then it goes through 2000-3000 pages for a typical film. Then it goes to the next film. (Did I mention that there are literally millions of films?)
However, when the documents are handwritten, and most Italian civil and church records are indeed handwritten, the shape of the letter “a” can vary based on the handwriting. Different clerks wrote records within the same town, and the “a” might be a little different. A left-handed writer will make the “a” different than a right-handed writer.
In addition, the lower-case “a” handwritten can look a lot like the lower-case “o.” It depends how the tail of the “a” or “o” swoops off the rest of the letter. If you have looked at Italian civil records and you have had a hard time knowing if the surname is “Menolascina” or “Menolascino” or “Menaloscina” or “Menalascino” because the “a” and “o” are similar, you understand what I mean.
As a human being with research experience, or just because you have cousins who spell the name “Menolascino,” you know what the name is supposed to be. A computer, because it cannot “think” the way we do, will objectively convert the handwritten letter into “a” or “o” without considering that somebody’s cousin spells it “Menolascino.”
If the computer just looks at the writing, and concludes that the name is “Menoloscino,” you may try to search for “MenolAscino” and not find the record.
So that is what “AI” computer technology is supposed to do. It takes all of the information in these documents into account when deciding whether the “a” is really an “a” or “o.” I do not know how it works but I imagine that after the AI computer reads thousands of pages within a specific town, and the name is “Menolascina” 75% of the time, then the “a” in the fourth character of the name is actually an “o” even if it looks like an “a.”
Now that I have confused just about everyone, what does this all mean?
A few weeks ago, someone told me that most of the records from Bari have been processed by the AI computers and were searchable on Familysearch “Records.” Until 2025, very few records could just be searched. You had to go into the FamilySearch catalog, find that town, find the births, find the right year, check the index (if there is one) and find the record. Sometimes, you would have to use Antenati Beni Culturali and do the same process. Now, you can go to FamilySearch and search by name, year, record type (Birth, Marriage, Death) and parents/spouse.
This has opened up the floodgates for me because of the number of families in my tree.
Why does it matter to search for the right names? Just go to the catalog and find who you need, right? Well, that system works best when the family stayed in one comune. If they moved around, or the parents came from two different towns, you might not find all the children in the same town. The solution is to search for the parents, and search the entire province or region and leave the town name out of the search, and you might trip over one child or more who was born in another town that you would have never looked in.
The other great feature is that you can try to search for deaths. Deaths are the hardest records to search for, because you don’t usually know when your Italian ancestor died. Let’s say your great-great grandfather was born in 1825. He married in 1848. His children were born between 1849 and 1862. He could have died any time between 1861 (maybe he died before his last child was born) and the 1920s. You would have to search each year of deaths between 1861 and 1925. Now, by searching FamilySearch, you can leave the year blank and just search for the name. The search results will show names of parents and the spouse, so you can be sure the one from 1883 is the correct one.
So far, I have found hundreds of additional children to add to many families, saving all the time of searching the “film” year by year. In many cases, the children were born in a “third” town. This means I have a father born in Triggiano and a mother born in Carbonara, and for some unknown reason they moved to Valenzano and their children were born there. Why would I just check Valenzano if the parents had no connection to that town prior to their marriage? I would have to check all the nearby towns and hope for the best. This would take a lot of time and I might come up empty. By searching the the “AI”-generated index, I tripped over the “third” town easily and found a bunch of descendants for a family I needed. Many families, actually!
I also found a lot of individual death records because I just searched for the name, or used wild cards like asterisks and question marks, and I also searched for either or both parents, or the spouse to narrow down the results. I have a lot more of these still to look for but this is going to save a lot of time. And as noted before, if the person died outside of their birth town, or outside of the towns their children were born in, I would have no way to find it without the index pointing to “third” town.
I started this column with the fact that there would be some bad news.
Let’s start with the obvious.
Bad news #1: The “AI” index is not 100% accurate.
Since we’re dealing with a computer, as opposed to a thinking human with research experience, the computer sometimes mixes up different letters. Lower case “a” versus “o.” Lower case “t” versus “l.” Lower case “e” versus “i” especially in names like De Nicolo or Del Medico. Lower case “q” and lower case “g.” Etc.
The same family may have children under “close” names. Example:
Vitantonio Mastrolonardo and his wife Maria Chiara Rubino
Some of their children show parents “Antonio Mastrolonardo” and “Chiara Rubino,” or “Vito Mastrolonardo” and “Maria Rubino.” Basically any combination of names could show up, which is the fault of the original records. But now that you are searching FamilySearch, you have to allow for multiple combinations of names, or you might miss some of the children of this family.
There are other mistakes I have noticed. I see a lot of people who were born in 1900 who were indexed as having been born in 1890. The AI should be reading the year as “Millenovecento,” which is in typescript, and conclude 1900.
Bad News #2: Not every Italian record has been indexed this way.
For Bari, it looks like they indexed the same “films” that were recently made available on Antenati Beni Culturali. https://antenati.cultura.gov.it/ Even though my town records go from 1809 to 1945, the records that were indexed only include 1809-1900, and 1930-1935. I wish I understood why. (As of this writing, the same records on “Antenati” have NOT been indexed.) You may look for all the children of a couple who married in 1890, and you will only find children who were born up to and including 1900. Your part of Italy may be different, so you should look for people you already know about, in order to see how many years were convered by the AI index.
Bad news #3: Not every record is readable by the computer
Yes I have been disappointed by this, because in some instances, I found children I did not know about, but I did NOT find children I do know about. Why aren’t they found? It could be that the index did not pick up the letters in the way they should. But the real reason might be that the page has ink bleedthrough, book worms, weak ink that is hard to see, and other reasons why even a smart AI computer cannot make sense of the shapes that appear on the page. As a result, that page may be removed from the index for further review.
In general, I give you the following advice. Use this index to lead you to the records that you need. But do not stop there. Go to the records themselves and use your “GI” Genealogical Intelligence so recheck the “AI” computer and see how accurate it is.
You may need to use Antenati to go to the actual record image – many images were removed from FamilySearch some time ago.
To get the low hanging fruit, as they say, search for the death record of an ancestor who died in the 1800s, and you do not have the death date yet. You might save yourself a lot of time searching for the death. And you might find the parents of this ancestor, if you didn’t have them already.
I am searching for Angela Pesole, so I use asterisks to cover for Angel”O” and “Pes”O”le
When searching for deaths, select the province from the place names. If you want to look in the specific town, that’s up to you, but I always leave the province open. If you just put “Italy” you might get too many results from towns and families that have no connection to you.
Click “Spouse” and type enough of the spouse’s name to narrow down all the Angela Pesoles to the one you want. You may have noticed I put a question mark in the spouse’s surname. His surname is “Lozupone” and lower-case “Z”s are particularly misread by AI. They come out as “g” a lot.
Click “type” and select only “Death.”
If you can’t find a marriage of one of your ancestor couples, try the same method as above but select “Marriage” and select either the Spouse or parents. You can search for marriages by bride or groom, but put the other spouse in the spouse area when searching.
Another search method is to find all the children of a couple who are your direct ancestors.
Same as above – click “Birth” and type in your birth location to search – I highly recommend using the province but you can do anything you want.
You need to search for the parents in order to find all their children. You might find similarly named parents from the wrong time frame but you can ignore those. In this example, it would probably be smart to replace “PAT*” with “P?T*.” The “a” could become “o” and you’ll miss “Potano.”
Again, click “Type” and this time select “Birth.”
This is a lot to process, and I didn’t want to split this into two columns and make you all wait a month longer. Try these methods and let me know if you solved an old problem using these methods. I cannot speak to how much of Italy was processed with the AI computers, or if other parts of the world were done the same way. You should try to search your area and write me at d.niemiec@comcast.net and let me know how you did.
