Evaluating the Efficiency of Automated Language Models in Processing Historical Manuscripts and Detecting Forgery:

A Comparative Analytical Study

Abstract

This study aims to determine the efficiency of Natural Language Processing (NLP) algorithms in identifying forgery in old texts. A specialized checklist was used on a sample of 48 historic manuscripts to evaluate five models: Tesseract, Kraken, eScriptorium, and Ai Studio. The research used an analytical descriptive approach. With superior overall performance in the areas of digital processing, fake identification, and text, the Transkribus model demonstrated The results show an 83.3% percent score since it depends on deep learning techniques created especially for historical scripts. On the other hand, the results of the other models in this field were below par. Establishing this paradigm in heritage institutions, developing large language models suited for Arabic script, and developing standardized databases to be used will all help to support the research. improving the accuracy of the models addressed in the study by training them.