Using Artificial Intelligence to Expedite The Research Process
Presentation Type
Poster
Presentation Type
Submission
Keywords
Artificial Intelligence (AI), Handwriting Recognition, Transkribus, Machine Learning, Jenny De Mayer
Department
History
Major
International Studies
Abstract
This project centers on the development and application of custom AI handwriting recognition models to support archival research in the humanities. We used the platform Transkribus to train and refine models to accurately transcribe a collection of handwritten, multilingual letters by Jenny De Mayer. De Mayer was a traveler from Russia whose writings span multiple countries and time periods. From 1885 until her death in 1951, she wrote regular letters to her family and friends detailing her experiences. The primary objective of this research was not only to digitize these documents but to evaluate how effectively AI can be tailored to interpret complex, individual handwriting across different languages.
Through training and testing, the models achieved increasingly accurate transcriptions and demonstrated the potential for machine learning tools to transform traditionally time and labor-intensive archival work. By the end of the summer, our handwriting-recognition models could read Mayer's handwriting in French, German, and English with a character error rate of approximately 5%. The resulting transcripts contribute to the creation of a digital archive, significantly improving accessibility and usability of historical materials. This project highlights how AI can be leveraged to expedite the research process, reduce barriers to working with primary sources, and open new possibilities for large-scale analysis in historical and humanities research.
Faculty Mentor
Sharyl Corrado
Funding Source or Research Program
Summer Undergraduate Research Program
Location
Waves Cafeteria
Start Date
10-4-2026 1:00 PM
End Date
10-4-2026 2:00 PM
Using Artificial Intelligence to Expedite The Research Process
Waves Cafeteria
This project centers on the development and application of custom AI handwriting recognition models to support archival research in the humanities. We used the platform Transkribus to train and refine models to accurately transcribe a collection of handwritten, multilingual letters by Jenny De Mayer. De Mayer was a traveler from Russia whose writings span multiple countries and time periods. From 1885 until her death in 1951, she wrote regular letters to her family and friends detailing her experiences. The primary objective of this research was not only to digitize these documents but to evaluate how effectively AI can be tailored to interpret complex, individual handwriting across different languages.
Through training and testing, the models achieved increasingly accurate transcriptions and demonstrated the potential for machine learning tools to transform traditionally time and labor-intensive archival work. By the end of the summer, our handwriting-recognition models could read Mayer's handwriting in French, German, and English with a character error rate of approximately 5%. The resulting transcripts contribute to the creation of a digital archive, significantly improving accessibility and usability of historical materials. This project highlights how AI can be leveraged to expedite the research process, reduce barriers to working with primary sources, and open new possibilities for large-scale analysis in historical and humanities research.