An application of array grammars to clustering analysis for syntactic patterns

A new syntactic approach for clustering analysis using array grammars is introduced. The distance between an array and a core grammar characterizing a class of patterns is defined. It turns out this definition of distance is more satisfactory than a direct measurement between two arrays through error transformations. A 2-pass clustering procedure is proposed. This procedure does not require 2-dimensional arrays to be encoded into 1-dimensional strings and it can obtain less confusion and more accurate results than some other methods in the literature. An example of classifying a set of English handwritten characters is illustrated. Finally, several interesting future research topics and open problems are discussed.