Automating Gravestone Documentation and Analysis With Generative Artificial Intelligence

Abstract

Online cemetery digitization systems are limited by the need for manual information extraction from gravestone photography and for content-area expertise in creating informative analyses. Due to these limitations, existing systems offer search-and-retrieve interfaces with limited context about the burial culture or history of smaller or less prominent cemeteries. Generative artificial intelligence (GenAI) can enhance such systems with automated information extraction, AI composed analyses and summaries, and interactive conversational interfaces to explore burial culture. We demonstrate the implementation of three such GenAI-based applications. First, we present a fully automatic multi-lingual information extraction pipeline that combines optical character recognition (OCR) libraries and Vision-Language Models (VLM). Second, we present a set of tools that automatically generate data-driven summaries and analyses of cemeteries and lists of notable burials within them. Third, we present a retrieval augmented generation (RAG) conversational interface that gives data-driven and fact-grounded responses about burial culture, community practices, and cemetery history. Together they enhance the user experience of online cemetery digitization systems, bringing them into the Web 4.0 world. We detail the implementation of all tools using the CemoMemo cemetery digitization platform, evaluate the OCR + VLM pipeline against ground-truth transcriptions, perform ablation testing to isolate each component's contribution, provide sample generated content from each tool on a collection of Israeli cemeteries, and provide the GenAI prompts used. The tools serve as models for how GenAI and Web 4.0 can enrich the understanding and appreciation of cemetery burial culture.