This repo contains a text fabric dataset of the Ugaritic text corpus. It is work in progress.
This dataset is developed as part of the CACCHT project, which is a collaboration of Christian Canu Højgaard, Martijn Naaijer, Martin Ehrensvärd, Robert Rezetko, Oliver Glanz, and Willem van Peursen. The goal of CACCHT is to prepare and publish ancient Semitic texts digitally, that can be used for research.
For this dataset, we cooperate with Tania Notarius (University of the Free State), Alex Sosnovshchenko and Kseniia Yerofeieva. Other contributors are Maria Simion, Lynn Strietzel, Ben Shields and Elijah Labowe-Stoll, volunteer assistants (Polis - the Jerusalem Institute of Language and Humanities).
Version 0.3.0 contains 342 independent KTU texts and three catalogue aliases, including 63 tablets newly available since 0.2.8. The additions include Ugaritic, Hurrian, a mixed Ugaritic/Hurrian text, alphabet exercises, and an unclassified fragment. RIH tablets are not included. Only completed reviews replace the reference texts; unfinished reviews are excluded.
The 0.3.0 release description lists every added tablet with its language and line count, explains every node, edge and feature, and reports changes within the previously published tablets.
The textual hierarchy remains tablet > column > line, with word nodes over
sign slots. New boundary nodes retain breaks and scribal rulings, and alias
nodes resolve catalogue redirects without duplicating their target text.
Invisible synthetic slots retain otherwise empty textual entities.
For queries, use line_label for a line's printed address, surface and
surface_cert for surface attribution, type to distinguish letters and lexical
words from gaps, numbers and editorial content, and emen, cert, alt and
cont for text-critical information. g_cons is a consonantal word reading;
sign/after and usign/uafter reconstruct the two text formats. The former
constant language=Ugaritic feature is retired because it mislabels Hurrian and
mixed-language texts.
Older TF versions remain available. Node numbers are version-specific; use tablet, column and line addresses when comparing releases.