Stylistic Authorship Comparison and Attribution of Spanish News Forum Messages Based on the TreeTagger POS Tagger

Electronic texts from emails, social networks or mobile phones are currently of interest in Forensic Linguistics. Many of these texts analyzed are well under 200 words long. This work aims at identifying text authorship by using part-of-speech tags over short texts. Our corpus consists of 28 texts taken from forum messages. The tokens of our corpora were annotated with parts of speech (POS) provided by TreeTagger. A frequency vector based POS features was created and the Euclidean distance among texts was calculated. Results show how 10 out of the 14 (71, 4%) test texts were correctly assigned to their author.