Stylometry-based author identification

Abstract

In this study, we examine three different stylometric measures for the author identification task, including vocabulary richness, word length, and parts-of-speech frequencies on a 34-opinion article dataset with nearly 3000 word tokens. The findings show that although each of stylometric measures is regarded as a feature representing authorial writing style, it on its own does not have enough discriminating ability. However, if these stylometric measures are combined with Correspondence analysis for parts-of-speech as linguistic variables, we can identify an author of a Vietnamese text succesfully.