StyleExplorer: A Toolkit for Textual Writing Style Visualization

摘要

The analysis of textual writing styles is a well-studied problem with ongoing and active research in fields like authorship attribution, author profiling, text segmentation or plagiarism detection. While many features have been proposed and shown to be effective to characterize authors or document types in terms of high-dimensional feature vectors, an intuitive, human-friendly view on the computed data is often lacking. For example, machine learning algorithms are able to attribute previously unseen documents to a set of known authors by utilizing those features, but a visualization of the most discriminating features is usually not provided. To this end, we present StyleExplorer, a freely available web tool that is able to extract textual features from documents and to visualize them in multiple variants. Besides analyzing single documents intrinsically, it is also possible to visually compare multiple documents in single views with respect to selected metrics, making it a valuable analysis tool for various tasks in natural language processing as well as for areas in the humanities that work and analyze textual data.