Engineered Features for Personal Textual Language Style Clusters
Mary E. Koone · 2024
Hand-crafted features to detect idiolect (individual language style) in short text passages with k-means clustering differentiate individuals based on stylistic indicators. With a relatively small number of input samples and low computational resources, results are comparable to larger-scale methods and are human interpretable. The features make use of common stylometry attributes. An entropy-based measure evaluates feature contribution.