AI-driven extraction and intelligent retrieval of missionary archives in Malabar: advancing preservation and accessibility with machine learning

Purpose This study shows how AI improves the transcription, indexing and searchability of historical documents by utilizing AI-driven Optical Character Recognition (OCR), Handwritten Text Recognition (HTR), Named Entity Recognition (NER), machine learning-based classification and transformer-based retrieval models. Design/methodology/approach This study uses a computational archival science approach to analyze missionary records in Malabar by combining machine learning-based text recognition, natural language processing (NLP), document classification and AI-powered retrieval models. Findings The findings show that AI and ML significantly improve the speed, performance and efficiency of archival digitization. OCR achieves up to 97.5% performance for modern printed texts, while HTR models exceed 92.5% for structured handwriting, demonstrating the efficacy of deep learning in text recognition. NER models successfully extract missionary names (91.3% F1-score) and locations (90.0% F1-score), whereas classification models such as Random Forest achieve the performance of 89.3% when categorizing archival documents, and bidirectional encoder representations from transformers (BERT)-based search engines scoring 93.5% Precision@10 and 91.2% Recall@10, demonstrating their superior ability to retrieve relevant archival records. Precision@10 means that out of the top ten retrieved results, 93.5% are relevant, while Recall@10 indicates that 91.2% of all relevant results were found within the top ten retrieved results. Originality/value This study presents a novel integration of AI and machine learning for the systematic extraction, classification and retrieval of historical missionary records, bridging the gap between historical preservation and computational intelligence.