A binary-tree approach to generating imaginary Chinese characters

Psycholinguistic tasks such as lexical decision and nonword reading require presenting participants with stimuli that follow the conventions of a language’s writing system, but that are not attested in that language. For alphabetic languages, these stimuli are known as pseudowords or imaginary words (Meara, 2012). Various approaches to generating pseudowords in alphabetic languages have been proposed, with different levels of sophistication (Duyck et al., 2004; Keuleers & Brysbaert, 2010; Testolin et al., 2015; Westbury et al., 2007).
Languages which represent words using ideographic characters, such as Chinese, require distinct methods. The radical-based approach (e.g, Sze et al., 2014) generates pseudocharacters by combining a radical from one character with the residual component from another. The stroke-based approach (e.g., Tsang et al., 2018) works by adding or removing one or more of an existing character’s strokes. However, these approaches do not take full advantage of the hierarchical nature of Chines characters. For instance, the residual component of a character is often made up of further subcomponents arranged in a certain layout (possibly again forming a standalone character).

As an alternative, we propose a method which describe characters by a binary tree, where non-terminal nodes contain information about lay-out and terminal nodes contain stroke patterns. Pseudocharacters can be formed by manipulating of these binary trees by insertion, deletion, or substitution of nodes (see examples). Using this approach, we can capture character hierarchy in the resulted pseudocharacters and manipulate character-likeness by manipulation of nodes at varying levels (see figures 1, 2, and 3). The binary-tree representation may offer substantial benefits for generating stimuli in psycholinguistics experiments involving Chinese characters.