This year’s edition of ICDAR will host the following tutorials. The exact schedule will be announced soon.
From Ink to AI: Innovations and Integration in Forensic Handwriting Examination
Jonathan Heckeroth, Zurich Forensic Science Institute, SwitzerlandElisa van den Heuvel, Amsterdam University of Applied Sciences, Netherlands
Samiah Ibrahim, ForensInk Corporation, Ottawa, Canada
Erich Kupferschmid, Zurich Forensic Science Institute, Switzerland
Angelo Marcelli, University of Salerno, Italy
Abstract
Despite decades of digitalization, handwriting continues to play an important role in civil and criminal proceedings, and forensic handwriting examination remains highly relevant in legal practice. At the same time, advances in pattern recognition and AI have produced automated methods that perform well on benchmark datasets, yet these tools have seen only limited adoption in daily forensic casework. This tutorial explores the reasons for that gap and examines how computational methods can be better aligned with the practical, legal, and methodological requirements of forensic handwriting examination.
The session provides an integrated overview of forensic workflows, the development of automatic handwriting and signature verification, and the historical relationship between forensic experts and computer scientists. It highlights current challenges including non-representative datasets, limited explainability, mismatched output formats, and poor integration into operational workflows. The tutorial also discusses practical pathways toward collaboration, benchmarking, validation, and tool development that support transparent, scalable, and scientifically grounded forensic decision-making.
Bio
Jonathan Heckeroth holds a Master of Science in Psychology from the University of Mannheim and has undergone additional training in forensic handwriting examination. After working for about ten years as a private expert in a German lab, he joined the Zurich Forensic Science Institute in 2024. Alongside casework, he conducts research on forensic handwriting examination and is pursuing a PhD at the University of Lausanne on digitally captured signatures.
Elisa van den Heuvel, PhD, is a forensic handwriting and document examiner specializing in cognitive psychology, fine human movement control, and execution. She trained at the Netherlands Forensic Institute and has contributed extensively to research on handwriting kinematics, anti-counterfeiting measures, and AI integration in forensic analysis. She has led international workshops and ICDAR competitions and continues to advance the field by combining traditional expertise with innovative technological approaches.
Samiah Ibrahim, BA, BSc, has been a forensic handwriting and document examiner for over 30 years and is co-owner of ForensInk Corporation in Ottawa, Canada. She trained at the Canadian Security Intelligence Service and has held management and forensic leadership roles within several Canadian federal agencies and major forensic professional societies. She is recognised for her international capacity-building work and leadership within the American and Canadian forensic science communities.
Erich Kupferschmid has an MSc in Forensic Science from the University of Lausanne and has worked as a Forensic Handwriting Expert at the Zurich Forensic Science Institute since 2010. He completed extensive training at the German Bundeskriminalamt and has contributed to research on digitally captured signatures and writing machines. He has also promoted the likelihood ratio approach and systematic forensic analysis procedures with strong emphasis on peer review.
Angelo Marcelli is Professor of Computer Engineering and Director of the Natural Computation Laboratory at the University of Salerno, Italy. His research focuses on neurocomputational modelling of handwriting learning and execution, automatic handwriting recognition, handwriting analysis for early diagnosis of neurodegenerative diseases, handwriting identification, signature verification, and historical document processing. He has published over 200 papers in international journals, books, and conferences.
Document Understanding with Neuro-Symbolic AI: From Traditional Parsing to GraphRAG
Thomas M. Breuel
NVIDIA Research
Email: [email protected]
Web Page: https://github.com/tmbdev
Abstract
This tutorial explores the evolution of document understanding systems, from traditional OCR and grammar-based parsing pipelines to modern neuro-symbolic architectures that integrate large language models with structured knowledge representations. It examines how combining neural retrieval methods with symbolic reasoning enables more robust and reliable document intelligence.
The session discusses key challenges in document processing, including error propagation in pipelines and limitations of purely statistical approaches. It also presents emerging architectures such as retrieval-augmented generation (RAG) and GraphRAG, highlighting their respective strengths and tradeoffs. The focus is on practical design patterns for building scalable and trustworthy document AI systems that can handle complex, real-world document collections.
Bio
Thomas Breuel works on deep learning, LLMs, document analysis, and computer vision at NVIDIA Research. Prior to NVIDIA, he was a full professor of computer science at the University of Kaiserslautern (Germany) and worked as a researcher at Google, Xerox PARC, the IBM Almaden Research Center, IDIAP, Switzerland, as well as a consultant to the US Bureau of the Census. He is an alumnus of the Massachusetts Institute of Technology and Harvard University.