Holographic manuscripts of lead authors have been sought after for centuries and represent valuable collector’s items, to the point of pushing the creation of forged ones. Not unlike false paintings and sculptures, forged authorial manuscripts have circulated in the antiquarian market and have been bought by Libraires, Archives and private collectors. The same situation applies to famous calligraphers, whose work was sought after and therefore forged and imitated. The phenomenon is widespread and concerns documents and authors from different periods and regions. Today, thanks to years of development and competitions in writer retrieval, we are a step closer to detecting these frauds in an effective manner. This would be useful for both scholars, antiquarians and collectors alike.

For most writers (authors and calligraphers), however, the task is complicated by three issues: 1. the small size of surviving documents; 2. the inconsistent quality of the digitisation process; 3. particularly for the modern period, writers use different handwriting styles: (a) formal for public texts; (b) informal for correspondence with friends and family; (c) very informal for drafts and marginal notes. Moreover, these handwriting styles evolve over time as writers age. All three issues are common problems with authorial corpora.

The challenges of the task reside, therefore, in:

a. Detection of falsification or imitation

b. Working with an uneven and relatively small dataset

This competition is inspired by the previous editions. The main inspiration is ICDAR2017 Competition on Historical Document Writer Identification, ICDAR 2019 Competition on Image Retrieval for Historical Handwritten Documents [HisIR19] for the overall methodology. There are yet additional challenges, which relate to:

• ICDAR 2015 Multi-Script Writer Identification and Gender Classification (MS-WIGC-2015), it shares the “multi-script aspect”. Here, not in the sense that one writer produces texts in English and Arabic, but each writer has at least two formally identified scripts in which they produce texts.
• ICFHR 2016 Competition on the Classification of Medieval Handwritings in Latin Script and ICDAR 2017 Competition on the Classification of Medieval Handwritings in Latin Script share the focus on the medieval and early modern periods - HisFragIR20
• ICFHR 2020 Competition on Image Retrieval for Historical Handwritten Fragments for the challenge of working with image patches at different scales. Here, in real-world conditions, the dataset will include full pages and fragments (either real or artificially generated by randomly cropping digital images).

This competition adds the challenge of calligraphic script and willful and skilful imitation of scripts that may be very formal. In this regard, the dataset differs substantially from the corresponding datasets and tasks in classical forensics competitions.

According to the XIA approaches, and in the interest of interpretability of the end results, an additional requirement of the competition is that participants provide explanations for the classification resulting from the processing.

We expect 7 to 12 participants.