|
30 Aug-4 Sep 2026 Vienna (Austria)
|
|
|
|
Competition outline and scheduleThe competition dataset will encompass three different subsets illustrating three case studies sharing the same challenging characteristics: diversity in the writer’s production, other writers purposely trying to imitate the authorial handwriting, and scarce and uneven data. The subsets are the following: 1. Christine de Pizan: scribes X and X’ and their colleaguesConsidered to be one of the earliest feminist authors, Christine de Pizan lived at the court of the king of France in the late 14th and early 15th centuries and wrote novels, poetry, and biography, her most famous work being The Book of the City of Ladies. Most importantly, she actively supervised the copying of her works, and many illuminated presentation copies are preserved. Scholars have identified three principal scribal hands, conventionally called X, R, and P, and their corresponding cursive accompanying handwritings X′, R′, and P′. These hands diverge significantly in ductus, letter morphology, and layout, yet are considered to represent the same individuals working in both formal and cursive modes. Writer X may in fact be Christine herself. Hand P is said to be highly variable and sometimes imitative of X (Ouy 2012: 23), while R has been identified with P. de Cruce (also P. de la Croix), a documented professional scribe (Delsaux 2011). The Christine de Pizan corpus comprises 54 author-supervised manuscripts (Ouy 2012), not yet fully digitized, complemented by approximately ten further manuscripts copied by scribes P and R that contain other contemporary works. Portions of this material have already served as an early study for automatic writer identification in historical scripts (Aussems 2013), but a systematic, open, and multimodal competition has never been organized. 2. Torquato Tasso holographic manuscript.Tasso (1544-1595) is considered one of the most accomplished authors and scholars of the Italian Renaissance. He had a very tragic life, including imprisonment in an asylum for many years, a detail that has always sparked interest, even beyond his masterful literary production, which was translated into many European languages during his lifetime. The interest, particularly lively during the Romantic period, led to the production of false holographic letters and poems that fed this obsession. These forged manuscripts have been included in his corpus for centuries, but scholars over the past 30 years began to spot them. Almost all his manuscripts have been digitised, including some of the forged ones. Main difficulties of this subset: uneven and very diverse handwriting, affected by age and precarious health conditions. 3. Yāqūt al-Mustaʿṣimī, beyond the schoolYāqūt al-Mustaʿṣimī (d. 1298) stands as the most regarded of scribes in Islamic calligraphy and among its most celebrated figures. Active at the turn of the late Abbasid and early Ilkhanid periods, he is credited with codifying the six canonical scripts that defined the aesthetic and functional framework of Arabic writing from the 13th century onward. His innovations shaped the later Ottoman and Persian calligraphic schools. His fame was such that forgeries of his works circulated from the medieval period well into the early modern era. The corpus attributed to him, with varying degrees of certainty, numbers around 140 manuscripts, dispersed across collections worldwide and often difficult to access. Produced within a rigidly codified tradition, these manuscripts leave little room for personal variation. Certain codicological features and signature patterns allow for a subset to be attributed with greater confidence. Yet no study has isolated the palaeographic traits of Yāqūt’s individual hand beyond a single manuscript; scholars have instead focused on the stylistic school he founded. It is time to move beyond this notion and study his hand as that of an individual scribe. The challenge is to automatically determine whether a manuscript can be attributed to Yāqūt or not, and to evaluate how closely algorithmic classifications align with codicological evidence. The competition is based on a crisp classification task. The dataset available will contain for each subset:
Because of the uncertainty that is at the core of all historical research, the test dataset will mix images with the above-listed labels, with images for which scholars have not reached a consensus, in order to evaluate the inter-competitor agreement on these images. This part will not be taken into account for the ranking (see below). Participants MUST provide:
Participants SHOULD provide an explanation for the classification process, namely, explain which script features influence the distance between images and the final classification. Competition organisers will interpret the inter-competitor agreement with regard to state-of-the-art scholarship on the historical sources. Schedule
|
