On a sunny day of 23 December 2019 magaghat.ai comes online after several months of intense research and development.
Build by mathematicians Dr Hayk Aleksanyan and Dr Ara Gasparyan, it creates innovative software based on modern approaches of artificial intelligence (AI) for researchers in Digital Humanities aimed at the analysis of ancient manuscripts. Our current focus is on classical Armenian language, also known as Grabar. Armenian is a separate and independent branch of the Indo-European language family, and Grabar is the earliest phase of the development of the written form of Armenian, that according to traditional classification spans the 5-11th centuries. Our flagship product, which is currently in the early beta version, is an image-to-text conversion software build to automatically read text from digital copies of ancient manuscripts written in classical Armenian. Differences of this quest from traditional OCR (optical character recognition) systems are substantial. To mention a few: the texts on the manuscripts might be degraded due to physical damage, there is no uniformity in character sizes, the lines of text not necessarily follow horizontal patterns, the writings are sometimes encoded, etc. Such differences necessitate a radically different approach for this problem than those followed by conventional algorithms. This makes the problem more challenging and certainly a lot more appealing to us, and its resolution sets up a new stage for more applications to come.
We are excited to see how ideas based on mathematics, computer algorithms, machine learning and software engineering open many doors for fascinating applications in the Humanities. While we have our agenda on what to bring next, it is one of our utmost priorities to understand the needs of the community better. Thus, if you are here and would like to share a thought, please do not hesitate to do so. We would be more than happy to hear what you think.