Abstract: The style authors follow to express their ideas has been a subject of great debate. Several perspectives have been followed to try to analyze the style. In this contribution we present a computational methodology to study the writing style in a collection of hundreds of texts. For each text several attributes, which include different time series, are extracted and a battery of tools from the signal processing and the machine learning communities are applied to identify a set of features that may define a candidate style space. We applied self-organizing maps to visualize how several authors are distributed in the high-dimensional space associated to the style, and to visually prospect the similarities between styles from different authors.
Loading