PhD project: Making sense of illustrated handwriting archives

Name: Mahya (M.) Ameryan

Supervisors:
Prof. dr. L.R.B. (Lambert) Schomaker
Dr. M.A. (Marco) Wiering

Summary of PhD project:

Accessing historical handwritten archives, cultural treasure from the past, involves difficulties. Aging manuscripts, different style of writing, weakly structured, ink problems and low quality of the material of papers make historical scripts hard to read. Moreover, sloppy writing under difficult field-work conditions make reading problematic.

Recently, many historical collections are scanned and stored in the digital systems. During recent decades, several methodologies for transforming images of handwritten scripts into letter codes have been developed. When, small lexicons or single-writer documents, such as address recognition and bank check processing, are considered, fairly good recognition rates are achieved. However, in investigating large lexicons and historical scripts, we face many challenges for which no satisfactory solution has been found thus far.

In this project, we aim to interpret the historical scripts and drawing collected by scientists of the Natuurkundige Commissie between year 1820 and 1850. It is one of the most precious collections within the Naturalis Biodiversity Center and contains 17000 pages of handwritten scripts and drawing of Natuurkundige Commissie’s scientific exploration of the Indonesian Archipelago. The manuscript consists the personal notes based of the scientists’ natural observation in the jungle in four languages including Dutch, English, German and Latin, and also drawings of some animals and plants pained by artists.

In this project, fundamentally new methods for handwriting recognition will be developed within the MONK framework, a state-of-the-art machine learning handwriting system. Apart from deep learning methods, multilevel knowledge will be used to integrate information on the committee’s voyages and other contextual knowledge of the species, anatomy, geographical locations and habitats to bootstrap and improve handwriting recognition.