Phenetic Approach to Script Evolution

Hosszú, Gábor László (2017) Phenetic Approach to Script Evolution. In: Kodikologie und Paläographie im digitalen Zeitalter 4. Schriften des Instituts für Dokumentologie und Editorik (11). Books on Demand Gmbh, Norderstedt, pp. 179-252. ISBN 978-3-7448-3877-1

Text cpda4_print_Hosszu_u.pdf Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Computational palaeography, as a branch of applied computer science, investigates the evolution of graphemes, explores relationships between scripts, and provides support for deciphering ancient inscriptions, among others. The author applied methods often used to describe evolutionary processes in phylogenetics to analyse the development of scripts. Unlike in the clear evolution of phylogenetics, graphemes used to describe the evolution of scripts are sometimes indistinguishable from their glyph variants. Moreover, the historical background is at times incomplete. In order to reduce uncertainty, the author developed an exploratory data analysis method that combines phenetic analysis methods with a cladistic approach. The paper details the tests the author developed to explore the relationships among 66 different scripts with 186 different features. To extract data for analysis required determining the similarity groups of glyphs and orthographical rules in different scripts; the input is data from humanities-based palaeography. Creation of the similarity groups of the glyphs is based on minimizing the differences between the topological properties of the glyphs and individual decisions in order to avoid homoplasies, as well as the erroneous omission of slightly differing but otherwise related glyphs. For the second purpose, the layered grapheme model and the concept of characteristic transformations of related glyphs were used. Based on the extracted features of the scripts, various machine-learning methods were applied, including multidimensional scaling, k- means partitional clustering, and various hierarchical clustering methods. These algorithms produced similar results, represented in two- and three-dimensional scatter plots and phenograms, which visualize the relationship between the scripts. These results roughly concur with the results of humanities-based palaeography; however, new conclusions can be also derived, including the introduction of the concept of witness scripts, and glyph- and grapheme- level reticulations, which are used to describe the possible relationship of graphemes and scripts. The presented results demonstrate the usefulness of a developed modified phenetic method in exploring the similarities of scripts, and based on the results obtained, some improvements in modelling the distribution of certain historical scripts were also proposed.

Actions (login required)

Edit Item