Authorship Analysis and Cross-Language Grammar Features
Capturing the essence of the writing style of authors is an important research area in natural language processing. It allows to identify and attribute the author of a previously unseen document, perform so-called style change detection (find the positions at which the author changes within a document), detect plagiarism intrinsically, develop new technology for writing support, or perform forensic analyses.
To date, detecting variations in the writing style belongs to the most difficult and most interesting challenges in authorship analyses. The task of authorship attribution is particularly challenging in scenarios where ground truth textual data is only available in different languages (for instance, for bilingual authors). Moreover, style change detection is the only means to detect plagiarism in a document if no comparison texts are available.
In our research, we focus on utilizing grammar features for several of the above-mentioned tasks. Thereby, we have pioneered work in cross-language scenarios, where authors have written documents in multiple languages. Current research in this field also covers the detection of social media bots, which have become a more pressing matter in recent years.
At DBIS, we are part of PAN, an international group of scientists focusing on the writing styles and habits of authors. The PAN initiative organizes shared tasks, where many researchers from across the world compete against each other in finding the best strategies to tackle problems in Authorship Attribution, Author Profiling as well as Multi-Author-Decomposition. Particularly, we are co-organizers of the Style Change Detection task at PAN.