📄 An Assortment of Articles in a Ugaritic Text

Summary

Watson re-examines a small Ugaritic tablet, KTU 5.23 (RS [Varia 23], AO 21.090). Its provenance is unknown, and an unskilled writer, probably a trainee scribe, cut it roughly. The text is a bare list of nouns with no surrounding sentences, so nothing in it can show what a word means. Only one earlier study exists, the first edition by Caquot/Masson (1977). Watson goes through all 19 lines. He compares the KTU reading with the first edition and proposes meanings from two sources: attestations elsewhere in Ugaritic, and cognates in Akkadian, Hebrew, Syriac, Arabic and occasionally Egyptian. Where a translation is in doubt, he marks it as such. His overall reading is that the list names household goods: oils, containers, textiles, tools and ornaments. They are not arranged in any visible order, and the ruled lines on the tablet remain unexplained. Many of the words occur only on this tablet. Watson therefore presents his glosses as conjectures, and he allows that some entries could be meaningless practice writing. The study matters because it supplies proposed meanings for rare words in the Ugaritic lexicon.

Key points

• The tablet is KTU 5.23, first published by Caquot/Masson in Semitica 27 (1977). It is beige and almost square, about 52 × 60 mm, and is written on both faces and the edges.
• Readings that differ from the first edition:
  • line 13 reads lḥm. tnṯ
  • line 16 reads mkkm
  • line 17 reads dqm, not dmq
• Some readings rest on well-attested words. Examples are šmn mr ("myrrh-scented oil") and šmn nr ("lamp oil"), both confirmed by other KTU texts and by Akkadian and Hebrew equivalents.
• Some readings are new proposals that go against earlier ones:
  • kdnt as "pitchers" (from Syriac), rather than the usual "she-mule"
  • qdḥm as "saucers", rather than Caquot/Masson's "briquet, étoupe"
• Words found only on this tablet include kdnt, mkkm, mskkm, qdḥm, tqnt, tnṯ and ṭrdy/ṭrdm. The sequence in line 7 cannot be interpreted.
• Watson finds no loanwords from non-Semitic languages, though some words may have been taken directly from Akkadian.
• The PDF header gives the date as Folia Orientalia vol. 47, 2010. The library metadata gives 2026.

Interesting ideas

• I1. Watson rejects a well-known sense of a word ("she-mule") when it clashes with the semantic field of the rest of the list (household items), and he applies the same test to "vine" for gpn; a list's shared domain can disambiguate the words in it even when there are no sentences. — For Mova (world): Score candidate WordNet senses of each item in coordinations and enumerations by how close they sit to the senses of the other items in the ontology, and log the competing sense as the explanation.
• I2. Each gloss comes with an explicit confidence level ("very tentative", question marks, "no more than conjectures"), and rival meanings are kept rather than dropped. — For Mova (world): Store lexical and ontology facts with a confidence tier and their alternatives, so later stages can show the reader how firm each inferred entity type is.
• I3. For hard spans, Watson considers that the scribe may have made a segmentation error, such as a misplaced word divider or one sign with too many wedges, or that a single string may be two separate words. — For Mova (tokens): When the lexicon cannot analyse a token, generate alternative splits and merges (for example, typos that join two words) as ranked hypotheses instead of forcing a single segmentation.
• I4. He keeps an attachment question open: dqm may modify the preceding noun ("two small carpets") or may be a noun of its own ("gauze"). — For Mova (tree): In terse, punctuation-poor list text, produce both the amod and the separate-item analyses and let frame or world plausibility choose between them, recording why.
• I5. Words that appear only once are explicitly flagged as possibly meaningless rather than forced into a meaning. — For Mova (learn): Use an explicit "unknown / possibly noise" outcome as a negative control in evaluation, and reward a model that refuses to invent meanings for unattested forms.