Emotional Fingerprint from Authors in Classical Literature

The Internet deeply changed the way people share their knowledge. Almost all content that people produces is now available in digital formats, like e-books, apps, newspapers, and magazines. That content has commonly some metadata available that can be used to generate complex recommendation systems that track content similarity. Since there is some effort in the literature to explore this direction, almost all use classical recommendation approaches, like collaborative filter data and information present on websites that sells books. While most efforts in the literature use features derived from the text syntax to create a recommendation model, our approach aims to trace an emotional fingerprint of authors extracted from their texts. This approach, known as psychometry, consists of the study of behavioral characteristics like positivity, negativity, sadness, fear, religiosity, sexuality, which are able to disguise individuals. Using two sentiment analysis lexicons and a collection of 641 books from the English literature written by 56 authors, we show the effectiveness of these psychometric features in order to trace those authors emotional fingerprint.