Convolutional Swin Encoder

A Unified Deep Learning Approach to Writer Attribute Prediction

Authors

  • Aditya Majithia City College of New York
  • Arthur Paul Pedersen The City University of New York (CUNY) https://orcid.org/0000-0002-2164-6404
  • Michael Grossberg The City University of New York (CUNY)

DOI:

https://doi.org/10.32473/flairs.38.1.138949

Keywords:

Authorship Attribution, Handwriting Analysis, Swin Transformers, multiple task learning

Abstract

This paper focuses on developing a deep learning architecture capable of identifying writers' attributes from their handwriting. It introduces Convolutional Swin Encoder (CSE), a novel architecture combining Visual Geometry Group Network (VGGNet) and Swin Transformer blocks. CSE is designed to handle multi-label classification using images of individual handwritten words. As a unified encoder, it can predict writers' attributes such as identity, gender, age, and handedness. Using a word-level segmentation approach, CSE achieves competitive performance compared to page-level methods, which typically rely on separate classifiers instead of a unified one.

Author Biography

Arthur Paul Pedersen, The City University of New York (CUNY)

Dr. Arthur Paul Pedersen is faculty doctoral lecturer in computer science at The City College of New York and faculty research scientist with the CUNY Remote Sensing Earth Systems Institute.  Dr. Pedersen earned his Ph.D. in Logic, Computation and Methodology from Carnegie Mellon University for his work in theoretical and applied probability and statistics and has since lead interdisciplinary research in positions at the Max Planck Institute for Human Development in Berlin, and Ludwig Maximilian University of Munich.

Dr. Pedersen conducts research at the interface of artificial intelligence and cognitive science, probability and statistics, logic and computation, decision and game theory, and network science and social networks. His current research focuses on fundamental and applied problems in natural language inference and human-machine interaction, theoretical and methodological foundations of artificial intelligence, and social network and economic modelling and analysis.  Current application areas include intelligence analysis and tradecraft, forensic expert testimony and reporting, handwriting analysis and authentication, and strategic information forecasting and design.

Driving the research efforts of Dr. Pedersen are theoretical and practical problems for reasoning under conditions of uncertainty, especially under conditions owing to incomplete or contradictory information, under conditions of unresolved conflict or competing interests, or under conditions of limited time, cognitive capacities, or computational resources.

Downloads

Published

14-05-2025

How to Cite

Majithia, A., Pedersen, A. P., & Grossberg, M. (2025). Convolutional Swin Encoder: A Unified Deep Learning Approach to Writer Attribute Prediction. The International FLAIRS Conference Proceedings, 38(1). https://doi.org/10.32473/flairs.38.1.138949

Issue

Section

Special Track: Neural Networks and Data Mining