📄 Geoffrey Khan (ed.): Semitic Studies in Honour of Edward Ullendorff

Summary

This is not a research article but a short German-language book review (Aethiopica 10, pp. 243–246; the journal header gives 2007 rather than 2012). Stefan Weninger assesses the Festschrift edited by Geoffrey Khan for Edward Ullendorff's 85th birthday. He concentrates on the Ethiopian-studies contributions and only briefly lists the rest, which cover Ancient and comparative Semitics, Hebrew and Judaica, Syriac and Neo-Aramaic, Arabic and Islamic studies, and Graeco-Semitica.

He welcomes several pieces. One is Huehnergard's explanation of why Geʿez diphthongs sometimes contract and sometimes do not. He also notes Knib's classification of an Ezekiel manuscript. He criticizes others:

• Zaborski cites Voigt's counterarguments but does not engage with them.
• Goldenberg spends too long on the ambiguity of the term "root".
• Hopkins blends genealogical and typological questions.
• Pankhurst's article had already appeared in Aethiopica.

Overall, the reviewer judges the volume a fitting tribute. For NLP the review offers no data or models, only methodological lessons about morphology, terminology and argumentation.

Key points

• Published by Brill in 2005 (vi + 367 pp.) as volume 47 of the series Studies in Semitic Languages and Linguistics, the reviewed book bears the title Semitic Studies in Honour of Edward Ullendorff. It opens with an addendum to Ullendorff's bibliography and contains 26 contributions.
• Huehnergard traces irregular contraction of aw/ay to o/e in Geʿez qatl nouns to the nature of the adjacent consonants.
• Zaborski argues that Geʿez stems 0₂ (qättälä) and 0₃ (qatälä) are independent bases rather than derivatives of the basic stem. He also rejects labels like "intensive" or "conative".
• Pankhurst's paper on Emperor Yohannes IV (1872–1889) duplicates one already printed in Aethiopica 8 (2005), 48–71.
• For a fuller discussion, the reviewer points readers to Ewald Wagner's review in Orientalistische Literaturzeitung 101 (2006).

Interesting ideas

• I1. Apparent irregularity can sometimes be explained by a conditioning environment, as Huehnergard did by looking at the neighbouring consonants instead of listing exceptions.
• I2. Treating derived-looking stems as independent bases, without fixed semantic labels, avoids forcing one meaning onto a morphological pattern; Voigt's objection warns against going to the opposite extreme of full lexicalization.
• I3. The term "root" covers two notions: a historical etymon and an abstract lexical core that becomes a word only when slotted into a pattern. Separating such senses prevents confusion.
• I4. Genealogical relatedness and typological similarity are independent axes, and mixing them produces pseudo-questions.
• I5. Citing a rival position without testing its examples leaves an impression of selective reading; real engagement means running the counterexamples.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

• I1 (tokens): When the lemmatizer meets irregular forms, first test whether surrounding phonological or orthographic context predicts them, and only then add exception lists.
• I2 (frames): Do not hard-wire a single semantic role shift to a derivational affix; keep a learned per-lemma mapping with a pattern-level default.
• I3 (world): Store distinct ontology nodes for homonymous technical concepts instead of merging them under one label.
• I4 (tree): When converting UD dialects or choosing auxiliary treebanks, record language family and structural similarity as separate features rather than one "closeness" score.
• I5 (learn): Every new repair rule should be evaluated on the counterexamples a competing rule was built for, as an explicit negative control.