New Full-Text Search at FamilySearch Uses Artificial Intelligence

New Full-text Search at FamilySearch Uses Artificial Intelligence

Last week at Rootstech, FamilySearch announced a new experimental feature to search several unindexed collections. This enhancement harnesses artificial intelligence (AI) and optical character recognition (OCR) to quickly search several valuable record collections. It has the potential to not only drastically cut search time, but it may also help break through brick walls.

Full-text Search is only available to FamilySearch account holders, but it is free to register/use. Researchers need to visit FamilySearch Labs and add the feature to their account. The search feature is experimental while they continue to improve its functionality and accuracy, but in my opinion, it works fairly well. Considering the option of a manual search, I’ll take experimental any day.

Currently, only two collections are available: 1) U.S. Land and Probate Records and 2) Mexico Notary Records. However, more collections are planned.

Why Full-text Matters
While some of these collections include an index, e.g., grantor/grantee for land deeds, these indexes only include those who bought or sold their land, for example. However, Full-text Search now enables you to find witnesses and other people mentioned in the deed like former owners, neighbors, or heirs. When using indirect evidence to formulate proof arguments, I think we can all appreciate the benefit in using a FAN Club approach (friends, associates, and neighbors). You can also search by location or geographic points of interest, and Full-text Search additionally offers filters for time frame, record type, record place, and collection.

Getting Started
It’s easy to begin. FamilySearch offers a simple set of instructions, which is to sign in to FamilySearch.org, go to FamilySearch Labs, and select the full-text feature. I also strongly recommend you watch their short 4-minute video which discusses how to use search term options to focus your results for a more expeditious and streamlined process. Trust me, it’s worth the 4-minute investment. The tutorial is accessible through FamilySearch Labs page.

Does it Work?
I spent the good part of the weekend using the new feature, and it’s already making an impact on my research. Indeed, I found one record in 30 seconds that I spent a good two weeks searching through unindexed microfilm collections. While there are some challenges in sorting through imperfect results or managing the filters, I have trust that it will only get better. Plus more collections are on the way. Here are some examples of what I was able to find.

Three Generations in one Land Deed
Through analysis of my autosomal DNA matches, I found an unlinked family cluster headed by Daniel McMasters (b. 1845) who resided in West Virginia but who was born in Pennsylvania. I believed him to be the son of John McMasters (1816-1861), but I had no document confirming it. Only indirect evidence.

Using FamilySearch’s Full-text Search with the search term “Daniel McMasters” and filtering it to Pennsylvania, I found 33 mentions across several documents. One was for an 1876 land deed where Daniel sold his share in his grandfather’s estate (John McMasters Sr.) in Philadelphia, which had been passed down to Daniel’s now deceased father, John McMasters Jr. An excerpt from the land deed is shown below. FamilySearch highlights your search terms on the document in yellow, and I’ve used red boxes to additionally highlight the mentions of Daniel’s father and grandfather.

It’s important to note that one search result is presented for each mention of your search term per image, not per document. So, in the above example, the land deed covered three document pages across two microfilmed images. Daniel McMasters’ name appeared on both images and so two results were returned for the one land deed document.

A Newly Discovered Family Member as a Witness
While again searching for my McMasters family, but this time in Greene County, Pennsylvania, I used the search term McMaster* and the filter for Greene County. The asterisk enabled me to search for McMaster, McMasters, or any other letters that follow McMaster. The search returned a couple of interesting results including a Larley McMasters, who provided testimony in 1825 in the probate record James Morrison (see below).

While I haven’t yet researched who Larley McMasters was, I do take note that James Morrison was enumerated next to Elizabeth McMasters in 1820 in Greene Township in Greene County, Pennsylvania. Elizabeth has been the focus of my past research and having a “new” individual to research may prove insightful for her identity. In fact, the Larley name may be a hint to Elizabeth’s surname or another allied family member. I never would have discovered, or easily discovered, Larley McMasters if were not for Full-text Search. Because Larley was a witness, he would not have been listed in the probate index and not easily discoverable unless you had reason to research James Morrison.

Concluding with Patience
As with every new research tool, we need to exercise patience. Character recognition is not perfect, but it is certainly better than trolling through microfilm after microfilm on your own. I find that about 10% of the results are errors. Mostly it’s character recognition, e.g., when searching for Larley, some results were for Darby. Some of the other errors are based on where the words are positioned on a page. For example, one image had multiple columns like a newspaper. When searching for William Wilson, the text search returned a match based on “William” in one column and “Wilson” in the next column, which was on the same line. From the computer’s point of view, William Wilson appeared spatially relevant. The actual document listed Lewis, William in one column and Wilson, Robert in another.

I thank FamilySearch for taking the time, effort, and expense in developing the new tool. For everyone else, happy hunting!


Subscribe to Blog Posts

 

Published by Rick T. Wilson, Ph.D.

As the Patternologist,™ I use advanced DNA tools and traditional records to solve genealogical problems. I have 30+ years of genealogical research experience, and I am professionally trained in the scientific research method.

Leave a Reply

Your email address will not be published. Required fields are marked *