Kokatsuji: A Visualization Approach for Typographic Forensics of Early Japanese Movable Type

TL;DR

This paper introduces Kokatsuji, a visualization system that aids typographic forensics of early Japanese movable type, combining machine learning and expert knowledge to analyze historical printing techniques.

Contribution

The system integrates an ontology-based visualization with machine learning to support detailed analysis of early Japanese movable type printing processes.

Findings

Revealed errors in segmentation and clustering.

Uncovered patterns of block reuse.

Supported expert hypothesis validation.

Abstract

We present a visualization system designed to support typographic forensics in the study of Kokatsuji, the short-lived tradition of Japanese movable wooden type printing. Building on recent advances in machine learning for block identification, our system provides expert users with an interactive tool for exploring, validating hypothesis, and integrating expert knowledge into model-generated results about the production process of early printed books. The system is structured around an ontology of four conceptual objects (spreads, segments, blocks, and characters) each corresponding to a dedicated view in the system. These coordinated views enable scholars to navigate between material evidence and computational abstractions, supporting close, near-by, and distant reading practices. Preliminary results from expert use of the system demonstrate its ability to reveal errors in segmentation, inconsistencies in clustering, and previously inaccessible patterns of block reuse.