Manual and Automatic Annotation of Late Antique Poetry: Six Years Later

• In Event: Every Tag and Tag: Annotating Hard-to-Note Phenomena in Jewish Studies

Abstract

At the 50th annual conference of the AJS, held in Boston, MA in 2018, I organized one of the
first sessions ever at the AJS focusing on Digital Humanities and Jewish Studies. (Later, in
2022, some of the lectures were published in a special issue of the Jewish Studies Quarterly.)
During the session, I presented an annotation project that was in its initial stages at the time.
The project involved manually tagging figurative language in late antique Hebrew liturgical
poetry using the annotation tool CATMA.
In this paper, I will present the current state of the project, which has evolved with the
introduction of additional tools and methods. These include a new visualization tool
developed specifically for the project ( visÀvis ) and computational efforts to automate text
annotation using machine learning methods, including recently developed Large Language
Models. The paper will not only reconsider the project's prospects at its inception but also
address the inherent tension between human and machine annotations and interpretations.
The discussion will delve into the theoretical aspects of these processes but will primarily
focus on practical issues surrounding text annotation, visualization, and the challenges faced
by both humans and machines in dealing with literary texts, as well as strategies for
addressing these challenges.

Author

• Ophir Münz-Manor, The Open University of Israel