[Paper Review] Artificial Intelligence in archival and historical scholarship workflow: HTS and ChatGPT

Salvatore Spina|arXiv (Cornell University)|Jul 5, 2023

Digital and Traditional Archives Management4 citations

TL;DR

This paper evaluates the integration of Artificial Intelligence—specifically Transkribus and ChatGPT—into archival and historical scholarship workflows, focusing on automated transcription, correction, and normalization of digitized manuscripts. Using 366 letters from the Biscari Archive, the study demonstrates that while ChatGPT produces some inaccuracies, AI-assisted normalization yields text that meets scholarly expectations, significantly enhancing data analysis and computational linguistic applications in historical research.

ABSTRACT

This article examines the impact of Artificial Intelligence on the archival heritage digitization processes, specifically regarding the manuscripts' automatic transcription, their correction, and normalization. It highlights how digitality has compelled scholars to redefine Archive and History field and has facilitated the accessibility of analogue sources through digitization and integration into big data. The study focuses on two AI systems, namely Transkribus and ChatGPT, which enable efficient analysis and transcription of digitized sources. The article presents a test of ChatGPT, which was utilized to normalize the text of 366 letters stored in the Correspondence section of the Biscari Archive (Catania). Although the AI exhibited some limitations that resulted in inaccuracies, the corrected texts met expectations. Overall, the article concludes that digitization and AI can significantly enhance archival and historical research by allowing the analysis of vast amounts of data and the application of computational linguistic tools.

Motivation & Objective

• To assess the impact of AI on archival digitization processes, particularly in transcription, correction, and normalization of historical manuscripts.
• To evaluate the effectiveness of ChatGPT in normalizing early modern Italian letters from the Biscari Archive.
• To examine how digitalization and AI tools can enhance accessibility and computational analysis of analogue archival sources.
• To explore the redefinition of archival and historical scholarship in the context of digital transformation and big data integration.
• To identify limitations and practical outcomes of using large language models in humanities research workflows.

Proposed method

• The study employs Transkribus for initial automatic transcription of digitized manuscripts from the Biscari Archive.
• ChatGPT is used to normalize the transcribed texts, correcting spelling, grammar, and linguistic inconsistencies.
• A test corpus of 366 early modern Italian letters from the Biscari Archive's Correspondence section is processed using ChatGPT.
• The output from ChatGPT is compared against expert-verified transcriptions to assess accuracy and reliability.
• The workflow integrates AI tools into a computational pipeline for historical text analysis, enabling large-scale data processing.
• Digital libraries and big data frameworks are leveraged to support the storage, retrieval, and analysis of digitized archival content.

Experimental results

Research questions

• RQ1 To what extent can ChatGPT accurately normalize early modern Italian manuscript texts from the Biscari Archive?
• RQ2 How do AI-assisted transcription and normalization processes compare to traditional manual methods in terms of efficiency and accuracy?
• RQ3 What are the key limitations of using large language models like ChatGPT in historical manuscript processing?
• RQ4 How does the integration of AI tools like Transkribus and ChatGPT transform archival and historical scholarship workflows?
• RQ5 Can AI-enhanced digitization workflows support large-scale computational linguistic analysis of historical archives?

Key findings

• ChatGPT produced some inaccuracies during the normalization of 366 letters from the Biscari Archive, indicating limitations in handling historical linguistic nuances.
• Despite errors, the corrected texts generated by ChatGPT met scholarly expectations, demonstrating practical utility in real-world archival applications.
• The integration of AI tools significantly enhances the efficiency of processing large volumes of historical manuscripts.
• Digitization and AI enable the application of computational linguistic tools to archival sources, expanding analytical possibilities.
• The study confirms that AI-assisted workflows can support scalable, data-driven historical research while requiring human oversight for quality assurance.
• Transkribus and ChatGPT together form a viable pipeline for automated transcription and normalization in archival scholarship.