Investigating the effects of recency and size of training text on author recognition problem

Abstract

Prediction by partial match (PPM) is an effective tool to address the author recognition problem. In this study, we have successfully applied the trained PPM technique for author recognition on Turkish texts. Furthermore, we have investigated the effects of recency, as well as size of the training text on the performance of the PPM approach. Results show that, more recent and larger training texts help decrease the compression rate, which, in turn, leads to increased success in author recognition. Comparing the effects of the recency and the size of the training text, we see that the size factor plays a more dominant role on the performance. © Springer-Verlag 2004.