Learn how to use Transkribus, ChatGPT and other AI tools to translate German church records, and other records, into English. Learn about cautions and practical uses.
This Webinar Short explores the use of artificial intelligence (AI) to aid in the challenging task of translating German church records and other genealogical documents into English. It demonstrates how AI tools can assist genealogists in deciphering old German script and provides a comparative analysis of different AI translation methods. While highlighting the potential of AI to streamline research, the webinar also emphasizes the importance of human verification for accuracy, especially with nuances in language and historical context.
To gain a deeper understanding of how AI is transforming genealogical research and to learn valuable techniques for translating German records, it is highly recommended to view the full webinar or consult the resources mentioned. The insights shared will empower genealogists to effectively utilize AI tools while maintaining the highest standards of accuracy in their research.
Transkribus will definitely help with my tougher records, those with less typical wordings. Thanks! In years past, I fed Tranlate output back as input in the other direction. Where it fails shows the areas where “nuance lossage” and “language aging” that Andrea mentioned, and also idioms, translational inexactitudes, and more are a problem. Feeding it backward allows you some freedom in carefully tweeking or massaging the inputs/outputs. Nice presentation.
I’ve used Transkribus and then DeepL. It is still amazing!
I wonder how Claude.ai would handle the transcript? It’s a game changer for English documents, so it would be interesting to see for German.
I have been using Transkribus to translate personal letters in German for a few weeks now, and I found very quickly that the results were much better if I paid to use the Super Model. I use Deepl to translate and the results are better if you edit the transcription so it doesn’t treat each line separately. It’s not perfect but the human alternative is not affordable and so it’s much better than never finding out what is in those letters.
When I finish this project I will transcribe my grandmother’s letters to my mother, another of those tasks I thought I’d never have time for.
Fascinating Geoff. AI is developing so quickly that I am rarely surprised by what it can do now… but quite honestly, I did not expect this!
I have a few Latin documents that I have attempted to translate but have not gotten very far. I think I’ll try your suggestion and see where it takes me. Thank you.
Did you also obtain the emigration papers for Catharina Kienz, Archiv B731/1 Nr. 902
http://www.landesarchiv-bw.de/plink/?f=5-2237644
Not yet, but she’s on my list. 🙂
I found this webinar fascinating and halfway through the proverbial light went on! I have a trove of about 20 letters sent to my great-grandfather that I have been “meaning to translate” for over 40 years. They are written in German script which is getting harder and harder for me to decipher over the years. I grabbed a four-page sample written (I suspected) by his sister in 1898. It took a bit to get it through Transkribus but I succeeded on my fourth try. Then I sent the output to ChatGPT and, oula, there was the letter in English from a neighbor of Louis’s sister who was very ill. Thank you so much for this webinar.
Wow, how wonderful!!
Many people who translate on a regular basis believe that DeepL does a better job of translating that Google Translate, particularly if it’s a longer document.