A Study on the Writer Identification Task for Paleographic Document Analysis

The subject of paleography is the study of ancient documents. In particular, the paleographer's aim is to locate a document in a cultural environment and chronological interval in the past. Automatic writer identification is then a desirable tool for a paleographer as she/he gains useful information about the document at hand. However, the paleographer is often interested in methods which can be easily interpretable by humans. In this paper, we apply some state-of-the-art techniques devised for modern documents to the paleographic domain. Moreover, we propose new techniques and document representations with the aim at producing more understandable representation of a writing style. Experimental results have been performed on a large dataset of paleographic images and demonstrate the feasibility of the proposed approach, and the suitability of this tool on helping the paleographer's work.