AI for Summarizing and Transcribing Documents

I listened to the Legacy Webinar, "AI and Genealogy: Trouble Ahead?", presented by Thomas MacEntee. He discussed using AI to summarize and transcribe documents. I tried this out, and it's really a great time saver! 

 

One of my grandmother's, Graciela del Castillo, was Nicaraguan. There isn't much published about Nicaraguan genealogy, and what's available. I was in contact with a Nicaraguan genealogy expert back in 2015. I've been going through his emails again as I work on my grandmother's line. He sent me a dozen emails with detailed information regarding Nicaraguan research. I was planning on summarizing the information and sharing it with a group I'm involved in. After listening to the webinar I decided to use AI to do it. Thomas MacEntee demonstrated three AI sites ChatGPT, Perplexity, and Gemini. I tried out all three sites. I love information presented in bulleted lists, and asked that the emails be summarized that way. 

 Here is one of the emails I asked for a bullet list summary for:

Here is what the three AI companies came up with: 

Perplexity came up with the most detailed summary. ChatGPT was in-between as far as details. Gemini provided the shortest summary. I like that Gemini provided an introduction in bold script (also liked being able to save the answers to my Google Docs). 

I decided to ask Perplexity and ChatGPT to provide a bold introduction to each bullet point. 

This is what Perplexity came up with:

This is what ChatGPT came up with:

ChatGPT didn't provide an introduction for each bullet point, as I asked. I like what they came up with anyway. Perplexity did exactly what I asked. 

I was about to get a pen and paper ready to start summarizing these emails. I'm glad I listened to the webinar first! I'm sure these summaries are better than what I would have come up with. 

 

A Couple Family Valentines

 

The only photo I have of my parents when they were dating in 1962

Going through old boxes I found some old Valentines.

This one below was from my father, Bob, to my mother, Edna, when they were dating in 1962. The front is made of padded cloth. The verse is very romantic! 

Another family Valentine I found has velvet on the front. Probably 50 some years old? 

Happy Valentines Day!

Automating the Genealogy Process to Save Time

Unless you enjoy spending long hours in front of a screen, automating genealogy tasks can free you from long hours analyzing DNA results, writing reports, and looking through your work for errors. There are people who work 40 hours, and more, a week who would benefit from more genealogy task automation. 

I really got to thinking about time saving tools after listening to a couple of great webinars at Legacy Webinars. "Explore Uncharted Paths in Tracing Brick Wall Ancestors Through DNA Network Graphs" and "Artificial Intelligence and Genealogy: New Beginnings in 2024" provide examples regarding how automation can save time. 

Analyzing DNA results is very time consuming due to the number of matches most of us have. Going through matches searching for shared matches is the best way to determine how we are related to someone. One method at AncestryDNA is using colored dots to sort matches into groups based on shared matching. It's easy to pick a match, click on them, and see which matches are shared in common and place color coded dots by their names. There might be a dozen shared matches. You can then click on these dozen shared matches individually and see matches you share with them. As you work through matches, and shared matches of matches, and their shared matches this gets very time consuming. Automating this process would save time. The companies could run a program to place color coded dots for shared matches or provide us with cluster charts. 

MyHeritage allows their customers to generate cluster charts from their shared matches.Their charts are quite simple, but helpful. Even more detailed charts could be produced by companies. If you watch the webinar about graphs you can see all of the possibilities. 

Family Tree software, such as RootsMagic, produce barebones narratives using our family tree details. AI sites like chatGPT can produce more polished readable narratives based on the narratives produced using your family tree software. You need to suggest a style to chatGPT to get an end product you like. Even then you may want to edit the wording, but it's still time saving. The more readable narratives can help you find errors and make details pop out you might have missed. Producing some of these narratives I found I had information in my tree I now disagreed with. 

I liked how this narrative about my del Castillo family came out using my RootsMagic information and chatGPT here's a sample:

"Through the centuries, through wars and revolutions, through migrations and milestones, the del Castillo family remained bound by a thread of shared history and shared destiny. Their story, etched in the annals of time, stands as a testament to the enduring power of family, love, and the human spirit."

I thought that was actually appropriate for this family. Sounds a little grandiose but the family was involved in the politics of Nicaragua, one ancestor serving as president during a civil war. 

This Moses Wray example is a little flowery for my taste, but the fact he died at age 75 jumped out at me more than it would in a barebones narrative:

"By 1786, Moses Wray's name echoed through the valleys of Franklin County, Virginia, marking his claim once more. But the winds of time whispered a somber tune, and Moses passed on before the autumn leaves could paint the hills in shades of gold. He took his final breath before the 4th of October 1802, at the age of 75, leaving behind a legacy woven into the very fabric of Bedford and Franklin County."

Ancestry is offering DNA customers an AI feature which they can use to get more information about their ethnicity predictions. It provides some background on the culture and region  someone is predicted to have ancestors from. 

An important use of AI is for OCR, which allows for faster indexing of records. Language translation is another very useful feature. 

I see a lot of uses for graphs and AI in helping to save time on tasks I'm not that interested in spending time on. Writing research plans, narratives, and other reports would be a good use of AI technology. I hope companies will see the advantages of new technologies and offer them to their customers. There are tools we can use right now ourselves, but other tools would have to be produced by DNA companies since we don't have access to the information to produce our own graphs of shared matches, for instance.