RLP 317: Transcribing Document Images with ChatGPT and Claude

Nicole (1s):
This is Research Like a Pro episode 317 Transcribing document images with ChatGPT and Claude. Welcome to Research Like a Pro a Genealogy Podcast about taking your research to the next level, hosted by Nicole Dyer and Diana Elder accredited genealogy professional. Diana and Nicole are the mother-daughter team at FamilyLocket.com and the authors of Research Like a Pro A Genealogist Guide. With Robin Wirthlin they also co-authored the companion volume, Research Like a Pro with DNA. Join Diana and Nicole as they discuss how to stay organized, make progress in their research and solve difficult cases. Let’s go.

Nicole (41s):
Today’s episode is sponsored by Newspapers.com. Hello everyone, welcome to Research Like a Pro and hi to mom

Diana (49s):
Hi, Nicole. How. are you doing today?

Nicole (52s):
I’m really good. I’ve been experimenting with the new ChatGPT-4 and just learning more about it. It’s been interesting.

Diana (59s):
I’ve been experimenting with that as well. In fact, we were talking before we started recording that I have been working on my titles for Roots Tech presentation proposals and using the chat bot to help me come up with some better titles because sometimes my mind goes blank when thinking of better words. So I was using the new ChatGPT-4 o which is the major upgrade, but then after about 15 or 16 interactions, it sent me back to the 3.5 version. So that was kind of fun just to test the limits and see what happened,

Nicole (1m 38s):
Right? Yeah, you can do so much now with the free version of ChatGPT, but there are limits to how many messages you can send in a short time,

Diana (1m 47s):
Right? It told me it would let me back in in about three hours. So you know I had been really desperate. I could have waited, but I just continued with the lesser version, which was totally fine for what I was doing. Well, our announcements for today, we have our Airtable guides available on the website. Airtable research logs for genealogy, quick reference, second edition, and tracking DNA matches, matches with Airtable, the quick reference guide. So if you’ve been thinking about Airtable wanting to try something new, these guides are really fabulous for getting you started and, and we’re excited for our next Research Like a Pro webinar which will be held on Tuesday, August 20th at 11:00 AM Mountain Time.

Diana (2m 29s):
And this will be presented by Steve Little. The title is Who’s Eli’s Daddy: A Civil War-era Open Secret – A DNA Case Study. So this sounds like a fascinating case. It’s about James Eli “Bawly” Bower who was born in 1863 in Ash County, North Carolina during the Civil War, or a family history suggests Bawly’s father was a confederate soldier who while on leave in 1862, allegedly returned home not to his wife and children, but to another woman, Margaret Riley Bower. Nine months later, Bawly Bower was born and shortly after the soldier, William McMillan was dead. This case study aims to determine if documentary evidence and DNA analysis, both autosomal and Y-DNA can confirm or refute the family legend that William McMillan is the father of Bawly Bower, born to Margaret Riley Bower.

Diana (3m 23s):
Well, that sounds so fascinating and this is a tricky time period during the Civil War North Carolina. So the topics will be North North Carolina, Civil War, Non-Paternity Event (NPE), Oral Family History, Y-DNA Testing, Autosomal DNA Analysis, Pedigree Collapse, Endogamy, Multiple Relationships, Visualization Techniques in Genetic Genealogy. Wow. We’re going to learn so much from Steve about how he tackled this huge project. Our next Research Like a Pro study group begins in August 28th and the early bird registration has ended, but there are still some spots available likely, so be sure you sign up for that if you’re interested.

Diana (4m 7s):
And we invite you to submit an application for being a peer group leader where you get a complimentary registration. So you might be a candidate for a peer group leader if you have gone through the course or the study group and written a report and feel like you can help a small group to follow the process. And we invite you to join our newsletter to learn what we’re doing and to get coupons for any specials that we have going on. All right, well let’s get to our topic today. Today we are excited to talk all about artificial intelligence tools and transcribing document images with ChatGPT and Claude.

Diana (4m 50s):
So artificial intelligence tools have made huge strides in transcribing handwritten text in recent years and we were also excited at Roots Tech when Family Search announced their full text search and that is using AI to transcribe deed and probate images and more. And I know when I was working on my past project, it was so nice to have that easily find deeds for my John C. Cline and then have the transcription right there. And it was a pretty good transcription. I was impressed with that. Well, today in this podcast we are going to share how to use ChatGPT and Claude to upload images in either the JPEG or PNG formats and then transcribe them quickly.

Diana (5m 40s):
These large language models that we will hereafter call LLMs do a pretty good job of reading clear handwriting and you can use them when the task is pretty simple And. it can take maybe three to four minutes, but you want to reduce that time to 30 seconds.

Nicole (5m 59s):
This is a good time to talk about some updates since I wrote a blog post about this topic. And as you’ve probably heard, artificial intelligence tools are changing rapidly all the time. Well, when I wrote this blog post, the paid version of ChatGPT was 4.0 and that was required in order to transcribe images because the free version ChatGPT-3.5 didn’t allow you to upload images to analyze. Well after May 13th, 2024 the free version of ChatGPT is called ChatGPT-4 o with a lowercase o, which stands for Omni, and that is free for all users.

Nicole (6m 46s):
So now you don’t need to have a paid version to do what we’re talking about in this podcast episode with ChatGPT and Claude by Anthropic, which is a break off of open ai, which makes ChatGPT. Claude is similar in its capabilities, but it had allowed free users to upload files. So some of the examples we’ll talk about are from Claude and some are from ChatGPT. And it’s great because now anyone can try this with just the free version on ChatGPT. While we’re on this topic, let’s just talk about what this new version of ChatGPT can do. It’s ChatGPT-4 o and it’s supposedly even better than GPT-4 and it’s available in the free tier, which is removing financial barriers for people and allowing more people to experiment and benefit from artificial intelligence tools, which is great.

Nicole (7m 41s):
And that’s something that Steve Little talked about in his blog post when this came out in May. So he has a great blog post about how ChatGPT-4 o is a game changer for free AI access. And he also talks about a possible handwritten text recognition advance in ChatGPT-4 o. And so it’s something exciting to experiment with. But he talks about how he tested the FamilySearch full text transcription of a deed and he transcribed the deed and then compared different versions of ChatGPT and the family search transcription with his to see how many errors each one had. And he found that the new GPT-4 o returned only nine errors while GPT-4 had 17 errors and FamilySearch’s transcription had 22 errors.

Nicole (8m 32s):
So it’s exciting that these chat bots are getting better and better and are more able to help transcribe handwritten text Before they were good at transcribing handwritten text, they’ve been good at transcribing printed text, so that’s always something that you can do. If you have a newspaper article or something that’s just printed and you want it to be transcribed, that’s a really easy option for you to get it transcribed. And that is ChatGPT -4 o.

Diana (9m 3s):
Wow, Thank you for leading us through the latest and greatest there. I think it’s fascinating that FamilySearch had 22 errors ChatGPT-4 had 17 and GPT-4 o only had nine. Isn’t that exciting that these chat bots are getting smarter and getting better,

Nicole (9m 22s):
Right? It is really exciting. And one thing I forgot to say about the new GPT-4 Omni is that it can reason across audio, vision and text in real time. So it’s supposed to be a lot faster. And it can accept input in any combination of text, audio, image and video. And it can generate any combination of text, audio and image outputs. So, it can do a lot and, it can do a lot more than its competitors. And some people think that that’s one of the reasons it got released in May was because of some other announcements by other competitors who are increasing their context windows like Google Gemini and allowing free users to do more things like Claude.

Nicole (10m 4s):
So this definitely puts ChatGPT ahead of the competitors.

Diana (10m 9s):
Competition is great in this circumstance. Well, let’s talk a little bit about this idea of using the chat bot to work with your documents because I think we have all seen documents that are multi-page You know that 36 page pension application or 50 page estate file, and sometimes you look at all of that transcription and you just, oh, I think it’s gonna take me weeks to get through this. Well, this might be a good place to learn how to use AI. So document images that work well with LLMs that you might wanna ask them to transcribe are typed texts or a mix of typed and handwritten text.

Diana (10m 52s):
So often we see short documents like this, a marriage bond and for example, you know you’d have the boilerplate language already typed in there and then there would be a line where someone, the clerk usually had to fill out the appropriate information, date and names and such. We could also see this in a birth certificate, a bill of sale, like I mentioned, a pension application. Often those pensions will have just regular typed questions and then somebody’s handwritten in all the answers. So in the blog post that Nicole wrote, there’s an example of a document that you could try to transcribe with the help of ChatGPT now 4 o, and it’s a parole document in a Confederate CMSR.

Diana (11m 40s):
So the Compiled Military Service Record where some of the document is printed text and some is handwritten. The other cards in the CMSR were manually transcribed into the Airtable research log, but this was longer and so having ChatGPT to transcribe that can save so much time and can be so helpful.

Nicole (12m 6s):
Yes, this was a Compiled Military Service Record for William T. Dyer and each of the cards in that compiled military service record usually would just say like muster just a few dates and a few notes. So it wasn’t very hard to just transcribe that quickly into my Airtable log, but because this one had a bunch of information typed out about the terms of being released as a prisoner of war and the parole and all that, I wanted to know what it said and to understand it. But I thought this might be a good chance to have these two paragraphs about that parole be transcribed by ChatGPT.

Nicole (12m 47s):
So a lot of the time we can use artificial intelligence tools to just save us a few seconds here and there, but it can add up. And that’s what Steve Little said a lot in our class, Empowering Genealogists with AI by the National Genealogical Society that you know every time you can save just a few seconds here or there with artificial intelligence over the course of a day, you’re saving you know more than just seconds it adds up and you could save yourself 20 to 30 minutes or more. So that was kind of the goal here, was to just save a few seconds, like instead of 60 seconds of me typing this out, it could take 15 seconds of working with chat GBT well before giving ChatGPT the job to transcribe the image.

Nicole (13m 30s):
It helps to provide it with some context. So I already knew who the person was and just in case ChatGPT can’t read the name, I thought I would just tell it the name in the prompt. So this is the prompt I gave it. You are an expert genealogist, which that’s something I usually use every time I’m doing genealogy and asking the chat bot to do something so that it activates the neural network that focuses on genealogy and has all of those boards ready. After telling it its role, then I said transcribe this page from the Civil War compiled military service record of William T. Dyer from Hawkins County, Tennessee. It says WT Dyer, a fourth sergeant of company D 31st Regiment, Tennessee volunteers, CSA and is signed William T.

Nicole (14m 14s):
Dyer. The Paroling Officer appears to be illegible but could be N Pullen of the 20th regiment of Illinois volunteers. So in my prompt, I kind of just gave it some of the info I could readily see. What I really wanted it to transcribe was all the paragraphs in the middle that were just kind of typed in, easy to transcribe. And I knew it would struggle with some of the handwritten text because sometimes handwritten text is hard to read. Providing some of that information is easy to do because I’ve already been looking at the compiled military service record and I know the name of the subject and I could guess at the name of the parole officer. So that’s what I did. So I could have just typed some of that basic info into my Airtable log, but I wanted to fully capture the essence of this document and I thought it would be better to do that by having a full transcription rather than an abstract or a summary.

Nicole (15m 2s):
And it actually would take me more time to try to make a summary of this document because I’d have to read the whole thing, understand it, and then condense it down to a summary. So that’s why I wanted to transcribe the full thing.

Diana (15m 15s):
Well that makes sense and I think it’s so great to think through the different uses and different scenarios will warrant different ways of using the AI, but something we always have to remember is to fact check. And that’s a very important next step in using these LLMs to compare the information from the generated transcription with the original. So in the parole document, the date at the top was transcribed as ChatGPT as July 5th, but the document actually says July 9th and the date at the bottom was transcribed as illegible. So you can prompt the chat bot to fix this and other things by saying the date at the top is July 9th then at the bottom it’s July 10th, add “N Poland”, quote in square brackets in the place of the illegible signature followed by 20th regiment, Illinois volunteers captain and Paroling officer.

Diana (16m 11s):
And then when the transcription is done, you can copy and paste the fixed transcription into your research log. So Nicole, do you think it would be more work to tell the chat bot to fix that or just to fix it yourself when you’re putting it into your research log?

Nicole (16m 26s):
I definitely think it’s faster to just tell the chat bot to fix it because it rewrites the whole thing correctly. So it’s less work for you to try to find it. You know you can just sit back and watch it do the work. You’re kind of like supervising your assistant. So you tell it to fix it, and then it also can remember things that you tell it to do. So by telling it to put an eligible name in square brackets, it’s remembering that that’s your preference And. it can hopefully remember that for the future and learn from that.

Diana (17m 1s):
That’s a great point. And I also wonder now that we have the option to just talk to it, you know, using audio rather than type it out. You could just be reading through it and talk to it. I mean, that’ll be fun to try.

Nicole (17m 14s):
Yeah, and I’ve been using the audio option on the ChatGPT app, And. It’s a great way to interact with the chat bot because it understands audio so well and can respond pretty easily to it.

Diana (17m 27s):
Great. Well let’s have a word from our sponsor Newspapers.com. Today’s episode is sponsored by Newspapers.com, your go-to resource for unlocking the stories of your ancestors, dive into the newspapers where your family’s history unfolds as you search nearly a billion pages in seconds. Newspapers.com offers an unparalleled treasure trove of historical newspapers providing a window into the past with papers from the 17th century to today. Newspapers.com is the largest online newspaper archive. It’s a gold mine for anyone seeking to uncover stories from the past. Whether you’re a seasoned genealogist or just starting your journey, Newspapers.com makes it easy to search for obituaries, birth announcements and the everyday stories that shaped your family.

Diana (18m 11s):
It’s like having a time machine at your fingertips. And here’s the best part, our listeners get an exclusive offer. Use promo code FamilyLocket it for a 20% discount on your subscription. That’s FamilyLocket at Newspapers.com. Sign up today at Newspapers.com and embark on a journey of discovery.

Nicole (18m 30s):
Alright, well the transcription with ChatGPT went pretty well even though it had a couple errors, but ultimately it saved me the time of those two paragraphs that I didn’t want to transcribe myself. It just talked about how you know, so-and-so is a prisoner of war in the hands of the United States forces and virtue of blah blah blah. And Vicksburg and the Garrison and this Lieutenant General and all that and it laid out you know the fact that he signs that he will not take up arms against the United States nor serve in any military police or constabulary force in any for garrison or field work, etc. So all of those different stipulations, it transcribes that all perfectly. And so that was great.

Nicole (19m 10s):
Well now let’s turn to another example and this one is using Claude by Anthropic. And Claude is a large language model as well, just like ChatGPT and you can go to it by going to Claude.ai. And in this example I simply uploaded a marriage license and the marriage license was from 1854 and was a mix of printed and handwritten text. And I asked the chat bot to play the role of an expert genealogist, then said, provide a transcription of this marriage record between William T. Dyer and Susan Webster. The clerk is James H.

Nicole (19m 51s):
Vance. And one thing I’ve found with the chat bots is that they do struggle with names because names are not as easy to figure out as words in a sentence that have context clues. So I usually just will read the names ’cause I can, I usually know them already or can easily figure them out. The result was that Claude gave me a few key details instead of transcribing the image. And in that it had a major hallucination, it added a made up detail that William T. Dyer was the son of Susanna Dyer. However, it, it followed that little summary with the actual transcription and in the actual transcription that it made, it gave the correct information and didn’t add anything.

Nicole (20m 33s):
So I thought that was really interesting that it decided to give some key details with something false in it. And then it gave the correct full transcription. So you have to watch these chat bots. ’cause sometimes there will be details that sneak in that are incorrect, but it’s just important, especially when we’re first starting to use them, that we fact check everything and just be really careful

Diana (20m 57s):
And maybe you can define hallucination for our listeners who are not familiar with that term,

Nicole (21m 2s):
Something made up. And it happens a lot with these large language models because they aren’t human, they’re just trying to predict the next tokens, the next parts of the sentence. And they are getting smarter and smarter. And the way that you can reduce the hallucinations is by giving them source information or text to work with and say, don’t add anything to this, just only use what I’m giving you. So that’s one way to get less hallucinations and less errors. But it is important to know that the more that you use this, you’ll see little errors because they just happen. So you gotta be aware of that. Well the transcription was really good. It starts with state of Tennessee Hawkins County to any regular minister of the gospel having the care of souls or any judge or justice of the peace of fore said county greetings.

Nicole (21m 50s):
James H. Vance clerk of the county court for the County of Hawkins sendth greeting. You are hereby licensed and commanded by me to solemnized the rights of matrimony between William T. Hawkins and Susan Webster by joining their hands together as husband and wife. Given at office in Rogersville the 21st day of November, 1854 and of American independence the 79th.

Diana (22m 12s):
I think it’s so interesting that they were tracking the year of American independence. I had love to see a study about when they quit doing that. ’cause I’ve seen that quite a bit in the 1800s. It’s kind of fun.

Nicole (22m 27s):
Yeah, so with the hallucination about adding in the name of Susanna Dyer, I told the chat bot, where did Susanna Dyer come from? The image doesn’t say anything about William T. Dyer being the son of Susanna Dyer, but the actual transcription you made was pretty good. The justice of the peace was actually Branch Tucker, so I didn’t finish reading it, but at the end it had James H. Vance clerk, which was correct. And then it had the date November the 23rd, 1854. This day the above couple was joined together as husband and wife by me. And then it had Burl Tucker, JP. Well, when I looked at the name of the JP, it was Branch Tucker. So I basically, I told it the two errors that I found. And it said, you are right.

Nicole (23m 7s):
My apologies. The marriage record does not mention William T. Dyer being the son of Susanna. I made an incorrect inference there and thank you for catching that. The JP’s name is Branch Tucker, not Burl. I’ve updated the transcription. And then it regenerated the transcription with those two things fixed. So I always do that. I like to teach it and have it fix it instead of editing it myself. I just think it’s easier to tell it the errors and then have it regenerate it. Then I just copy and paste the final result into my research log.

Diana (23m 39s):
Okay, great. Let’s talk a little bit now about using another tool called SnagIt to transcribe newspaper articles. So what can you tell us about that?

Nicole (23m 50s):
Well, in Steve Little’s NGS course Empowering Genealogists with AI, I learned about a tool that he uses called SnagIt and it’s available at techsmith.com. This tool takes screenshots and I’m sure a lot of people use it who are listening. But it takes screenshots. And then one of the cool features of it is that it allows you to grab text from the image. So if you take a screenshot of a newspaper article or if you take a screenshot of a page that has typed text, not handwritten, then it can grab the text. So this is not an AI tool, but it can use OCR to grab text from non-handwritten materials. And I have used it a bit over the last few months and I really appreciate the ability to quickly grab text to paste into my research log.

Nicole (24m 34s):
One thing I don’t like about it is that it doesn’t add nice line breaks or formatting like ChatGPT and Claude. Actually what it does is it adds too many line breaks. I don’t like the line breaks, I want it to just all flow correctly. One of the strengths of ChatGPT and Claude in these chat bots is their ability to format text nicely and use language well. So, it does have that ability to smooth things out and make it look nice. So you have to kind of balance that with the problem of the hallucinations and the errors added in. Well, I found a newspaper article and took a screenshot of it with SnagIt and then had it grab the text And. it did a pretty good job, but it, it does have a lot of errors in it actually.

Nicole (25m 15s):
And you can see in the blog post that it has errors, you know capitalization errors. Sometimes in those old newspapers you’ll see like some words are faded, they’re on a fold line, you can’t read it very well. And OCR really struggles with those. And so there’s probably like 15 or 20 errors in this five paragraph newspaper article. When I put it into SnagIt instead, it did a much better job And, it didn’t have any errors at all because this is one of the strengths of the chat bot is that it can predict what the next word is. So even if it has a hard time reading it, it can guess really well what the next word should be. so it did a lot better.

Nicole (25m 58s):
So for example, in the last paragraph it says, Johnny Dunbar arrived yesterday with six runners and Mr. Cole of Pittsburgh, Kansas came in from Kansas with five horses. So ChatGPT got that perfect and then SnagIt with OCR grabbing the text, it put Johnny spelled JC instead of JO because it couldn’t read the O very well. So that was an error. And then instead of yesterday, it had broken yesterday across two lines and so it had yes space comma space terday. And then instead of with it had WLH, because there was just a little faded spot there, the word wasn’t easy to read. And then for Pittsburgh it had an L instead of an I.

Nicole (26m 40s):
So there were in that sentence like four things that I would’ve had to fix manually. So, it really works well for newspaper clippings where it’s just printed text, but for whatever reason you want to put that into a Word document or something in your report and you just want a transcription of it, which I usually like to have for my research log. It’s just really convenient way to do it.

Diana (27m 3s):
So did you take the snippet and upload that or how did you do the specific getting the clipping into the ChatGPT?

Nicole (27m 13s):
Oh, great question. Yeah, I just took the screenshot and pasted it in. So that’s one of the things that it’s so easy and fast with ChatGPT, is that you don’t have to even upload it. You can just take a screenshot with SnagIt or with your screenshot software on your computer. And then once you do that, usually it’s in the clipboard for you to copy and paste. And so then I just go to ChatGPT, click in the little box where you message ChatGPT, and I do Control V to Paste and then it’s there.

Diana (27m 47s):
That’s fabulous. That’s so great.

Nicole (27m 49s):
It’s really fast. And that’s one of the things I love about it is that it just takes away a lot of the steps for having to transcribe something. It it really is fast.

Diana (28m 0s):
That’s wonderful. Let’s look at a couple more examples of transcription with ChatGPT. You know, hopefully this will give you some ideas for using ChatGPT and Claude to help make your transcription tasks more efficient. One of these examples, it was a marriage record and the prompt was you are an expert genealogist transcribe this marriage record for Richard Dyer and Highly Couch. The Bronzeman is William Everhart. The eyewitness was EW Hedrick. This was one of those records that has type texts as well as fill in the blanks for the different people involved. So that came through fairly decent. And then the second example was an abstract from a Charles County will book, and this one was all typed.

Diana (28m 45s):
And it had a lot of names in it. It said you are an expert genealogist transcribe this will abstract from the Charles County will book number 16. And looking at the image on Nicole’s blog post, it looks like it did a fairly good job of getting all those names correct and, you know, typing names if you’re doing that yourself could be kind of a pain, but how nice to have it just do it for you. I noticed that for Permillia instead it said Pamilla. So You know every once in a while get something a little off. But this example really was pretty spot on with all the names, having them all spelled correctly straight from that little clipping of the will books.

Diana (29m 30s):
So that’s fabulous.

Nicole (29m 32s):
Yeah, it did get Pamilla instead of Permillia. And that’s funny because that’s not a name you ever hear anymore, but back when a lot of our ancestors lived in the south, in Texas and Tennessee, we see that name all the time.

Diana (29m 44s):
Yeah. So at least per, we’re like, well that’s obviously Permillia

Nicole (29m 50s):
Right. Yes.

Diana (29m 51s):
So funny, that just goes to show the experience of reading a lot of records and working with a lot of old time names.

Nicole (29m 57s):
Yeah, I was gonna say with that marriage record for Richard Dyer and Hyley Couch that it transcribed Hyley’s name with HI, even though in the record and in my prompt, I spelled it with the HY. So I was trying to like prevent it from spelling the names wrong by giving it the correct spelling in my prompt. And so I spelled it H-Y-L-E-Y. But then when it did the transcription, it still said, you know, the marriages said Richard Dyer to Hiley spelled with an I and then with square brackets, Hyley spelled with the Y. Mm.

Diana (30m 30s):
That’s funny. So

Nicole (30m 31s):
That was interesting. And, it did follow my chosen convention of putting illegible things in brackets with a question mark. So it didn’t really try to read the year. The year was completely legible. It was 1858, the eight was legible, but the printed part was 1850 blank. And so then they had to fill in what year it was So, it was 1 8 5 blank. And then ChatGPT just put it in question mark like I’m not even gonna try to read that here. Like, come on, you can do this. Yeah,

Diana (31m 3s):
That’s funny.

Nicole (31m 4s):
Yeah, so while it does save some time, it’s not error free and we have to check it. Hopefully this was helpful to kind of see kind of where things are at. And all of these examples were done previously to the release of ChatGPT-4 o. So in Steve Little’s blog post, he shared some tentative information that the new 4-o is a little better at this handwritten text recognition. So it’s getting better, but it’s not gonna be perfect. We have to fact check, but I think it can save us time and I think it’s good to try it out, especially if you have these type of images that are mixed between type text and handwritten text and you would like to have a full transcription for your research log or for your report.

Diana (31m 46s):
Yeah. And just to make it easy for you to read through it. Sometimes it’s just kind of difficult to read in the old typeface just because it’s small it our eyes sometimes don’t wanna focus on it, at least mine don’t. And so just to have it in normal type and you can make it as big as you want and you can read it and try to understand it because the whole value of doing this work is to get every detail out of these documents possible. And we just wanna get the technical part out of the way so that we can then use it in our research, save us some time in one place so we can spend more time on the analysis and using it.

Nicole (32m 23s):
Absolutely. And checking for errors does help you analyze it more fully. You know So, it’s kind of like rolling some of that analysis into the transcription piece.

Diana (32m 33s):
Exactly. Because that is part of the value of doing a transcription yourself. And we have talked a lot about this, that there is value to reading every single word of a will or a deed to pick up on everything. And so this could just make it a little bit easier and then we could do the fact checking to still get all of that internalized, all those details.

Nicole (32m 53s):
Right. We don’t want to give so much work to the chat bot that we aren’t understanding the records anymore because we are not even looking at them.

Diana (33m 2s):
Right.

Nicole (33m 3s):
Well, we just finished our Research Like a Pro with AI workshop, and if you would like to access those recordings, you can still purchase them on our website. So go ahead and go to our website to learn more FamilyLocket.com/shop and go to the Artificial Intelligence category to learn more about this workshop. You can have access to the recordings till the end of August.

Diana (33m 26s):
Well, we hope you have enjoyed listening to this podcast and we hope you are on the journey with Nicole and myself to discover more about how AI can help us in our genealogy. So thanks everyone for listening and we’ll talk to you next time. Bye-Bye

Nicole (33m 42s):
Bye. Thank you for listening. We hope that something you heard today will help you make progress in your research. If you want to learn more, purchase our books, Research Like a Pro and Research Like a Pro with DNA on Amazon.com and other booksellers. You can also register for our online courses or study groups of the same names. Learn more at FamilyLocket.com/services. To share your progress and ask questions, join our private Facebook group by sending us your book receipt or joining our courses to get updates in your email inbox each Monday, subscribe to our newsletter at FamilyLocket.com/newsletter. Please subscribe, rate and review our podcast. We read each review and are so thankful for them. We hope you’ll start now to Research Like a Pro.