English. This paper describes the addition of an index of 1; 763 Ancient Greek loanwords to the collection of Latin lemmas of the LiLa: Linking Latin Knowledge Base of interoperable linguistic resources. This lexical resource increases LiLa's lemma count and tunes its underlying data model to etymological borrowing.
“Graecia capta ferum victorem cepit”1
HORACE, Epistles, II, 1, 156
Boasting over two thousand years’ worth of
written attestation, Latin’s evolutionary history is
among the longest in existence. The diachronic
and geographical reach of the Roman Empire
exposed Latin, an Indo-European Italic language, to
many regional dialects and languages, including
Ancient Greek. The mutually profitable linguistic
contact between Latin and Ancient Greek2,
facilitated by their similar morphosyntactic structures
and characteristic syntheticity
Examples of Latin words calqued from Ancient Greek are unicornuus “unicorn” (unus “one” + cornu “horn”) from monìkerwc (mìnoc “one” + kèrac “horn”), and infans “infant” (in- “not” + fans “speaking”) from n pioc (negative prefix nh
Copyright c 2020 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
1“Captive Greece captured her savage conqueror” (our translation).
2In Egypt, for instance.
+ êpoc “speech”). Calques can also involve
affixes, as is the case of Latin’s suffix -us being
substituted for the Greek -os
Example loanwords in Latin are crocodilus “crocodile”, imported from the Ancient Greek krokìdeiloc, and liquiritia “liquorice” from glukÔrriza. Adams identifies three categories of Greek loans in Latin (2003, p. 443): (1) words for which there existed a Latin equivalent; the writer was so familiar with the local Greek term that he adopted it in response to local conditions; (2) local Greek technical terms for which it might have been difficult to find a Latin equivalent; and (3) transfers determined by a writer’s lack of fluency in Latin, as a result of which he either adopted Greek words because he was unaware of their Latin equivalents, or did so unconsciously because of his poor command of Latin.
For each category, Adams provides a handful of examples, including (1) (h)amaxa from maxa “wagon”, (2) buneurum from boÔneuron “whip of oxhide” and (3) arura from roura “land”.
Over the course of its long history, Latin
lexicography has produced a plethora of lexical
resources, notably dictionaries, thesauri and lexica.
Many are available in machine-readable form but
their differing annotation schemes and formats are
seldom interoperable. In an effort to offset the
issue, the LiLa: Linking Latin project is leveraging
Linked Data technology to dovetail a wide range
of Latin resources into an interoperable whole,
producing an ever-growing lexically-based data
model capable of accommodating etymological,
morphological, syntactic and semantic
information, and more besides
Though chiefly targeting readily available
lemmatised resources on the web, LiLa also creates
linguistic resources in-house as a means of further
developing its underlying data model. Examples
of these are the Index Thomisticus Treebank
Etymological data is not new to LiLa.
Mambrini and Passarotti (2020) describe the inclusion
of 1; 391 entries from the Etymological Dictionary
of Latin and the other Italic Languages
The Index Graecorum vocabulorum in
linguam Latinam translatorum quaestiunculis
auctus (hereafter IGVLL) is a list of 1; 763 Ancient
Greek loanwords in the Latin language published
in 1874 by classical scholar Günther Alexander
E. A. Saalfeld. An extended edition of the
Index, published in 1884 as Tensaurus Italograecus:
Ausfürliches historisch-kritisches Wörterbuch der
Griechiscen Lehn- und Fremdwörter im
Latenischen, is the most comprehensive lexicographic
3https://lila-erc.eu/
4The Latin verb graecisso¯ used in the title of this paper
is a nativised version of the Greek graikizw “to imitate the
Greeks; speak Greek”. For a detailed overview of linguistic
“nativisation”, see Hock and Joseph (2009, pp. 247-57).
collection of its kind, counting roughly six to eight
thousand entries
Of the two, the size and Optical Character Recognition (OCR) quality of the 1874 edition best suited a first development of a derivative linguistic resource, conducted as part of a Master’s internship at the CIRCSE Research Centre in Milan6.
IGVLL is structured into three columns of information: the Latin loanword (occasionally accompanied by variants), the Ancient Greek source lemma(s) (multiple lemmas include graphical, morphological and dialectal variants), and a record of attestations (see Figure 1). Explanatory notes at the bottom of the page provide additional context. In thirteen cases, question marks indicate some level of uncertainty7, and, as is convention, asterisks are used to identify thirty-nine unattested –and thus reconstructed– Ancient Greek forms.
The OCR quality of the text written in the Latin alphabet, however, was sufficient to automatically isolate and tabulate the Latin lemmas, which were then manually cleaned. Next, this list was automatically mapped against the LiLa lemma bank to measure the degree of lexical overlap, which came up at 1; 488 unique matches (84:40%), 207 5Crude estimate of an average ten to fifteen entries per page, for a total 592 pages.
6https://centridiricerca.unicatt.it/ circse_index.html
7The specific uncertainty remains unclear as no editorial documentation is provided. ambiguous matches (11:74%) and 68 unmatched lemmas (3:85%). Unique matches inherited their respective LiLa identifier, ambiguous matches were manually disambiguated, and unmatched lemmas were added –once again, manually– to the LiLa lemma bank. Ambiguities were caused by homography between lemmas belonging to different categories, be those morphosyntactic (the lemma philosophus, for instance, matched against LiLa’s adjective philosophus “philosophical” and common noun philosophus “philosopher”) or inflectional (the common noun er might refer to the masculine er, eris “hedgehog” or the invariable er (graphical variant of R) “seventeenth letter of the Latin alphabet”). Of the 68 unmatched lemmas, 33 were graphical variants of lemmas already present in LiLa and 35 were new additions.
Next, we OCR’d IGVLL with Tesseract v. 4.1.1
set to Ancient Greek recognition
In LiLa, a lemma can have one or more graphical variants, known as “written representations” (e.g. the verb sacrifico “to sacrifice” is also attested as sacrufico), as well as inflectional variants, with which it holds a symmetric “lemma variant” property or relation in the Knowledge Base (the active sacrifico, sacrufico vs. the deponent sacrificor, sacruficor).
Therefore, for the purposes of LiLa, where the editor provides multiple Ancient Greek lemmas for a single Latin loanword, e.g. burrus “red” purrìc (pursìc); cyperum “rush (botany)” kÔpeiron (kÔpeiroc), these were distinguished into written representations of the same lemma (i.e. purrìc vs. pursìc) and lemma variants (i.e. the neuter kÔpeiron vs. the masculine kÔpeiroc).
Compounds such as authepsa “an urn, boiler”, 8For the most recent overview of Ancient Greek optical character recognition, see Robertson and Boschetti (2017). (aÎtìc & éyw) were tabulated as two separate words, and entries followed by a question mark (13 in total) were marked as “uncertain”. 2.2
Data Model The transformation of IGVLL into an RDF lexicographic resource bound for LiLa relied on a combination of vocabularies. In line with previous etymological work, we integrated the aforementioned lemon and lemonETY modules of OntoLex to represent lexical entries in IGVLL. The example lemma abacus “sideboard” shown in Listing 1 is treated as an ontolex:LexicalEntry linked to LiLa’s own abacus (lemma ID 86829) through the property ontolex:canonicalForm. a ontolex:LexicalEntry; rdfs:label "abacus"; ontolex:canonicalForm <..lemma/86829>;
Listing 1: Latin
We employed the Simple Knowledge
Organization System (SKOS)
[ontolex:writtenRep ">bax"]; skos:exactMatch <urn:cite2...:n51>.
Listing 2: Ancient Greek
The skos property stores the LSJ identifier of bax as an exactMatch to denote an exact correspondence between the Ancient Greek lemma of IGVLL and that of LSJ. Failing an exact match, the property skos:broadMatch is used to indicate that the IGVLL lemma is incorporated in a different entry of LSJ (e.g. the IGVLL noun fusik “science of nature, physics” does not have its own entry in LSJ but is listed as a nominalised adjective under the adjectival entry fusikìc “natural”); further, failing both exact and broad matches, the property skos:relatedMatch is used to indicate a loose relation between IGVLL and LSJ (e.g. IGVLL’s porfurÐzon “purple dye pigment”, neuter present participle of porfurÐzein, and LSJ’s verb porfurÐzw “to be purplish”). As LSJ is not currently equipped with a URN resolver, no actionable link can be made between LiLa and LSJ.
If multiple written representations of a Greek word are listed in the IGVLL, those are all assigned to the canonical form of the related etymon, for instance ontolex:canonicalForm [ ontolex:writtenRep "purrìc", "pursìc" ].
In the case of multiple Ancient Greek variant lemmas, these are all treated as individual etyma, with the difference that the primary etymon points to the URI(s) of the other etyma –classed as both lemonEty:etymon and lemonEty:cognate– via the additional property lemonEty:cognate (Listing 3). a lemonEty:etymon; rdfs:label "kÔpeiron"; lime:language "grc"; ontolex:canonicalForm
[ ontolex:writtenRep "kÔpeiron" ]; skos:exactMatch <urn:cite2...n60988>; lemonEty:cognate <http://lila.../IGVLL/id/etymon/499>. Listing 3: Lemma variants: kÔpeiron/oc Latin composite words in IGVLL never point to an Ancient Greek compound but to the two constituent lemmas. In contrast, in the LSJ lexicon seven of the total thirteen multi-word lexical entries in IGVLL are traced back to a Greek compound lemma, e.g. authepsa (IGVLL: aÎtìc & éyw; LSJ: aÎjèyhc9). In keeping with the IGVLL, we employed the decomp:subterm property of lemon10 to point the Latin lexical entry to its two constituent Ancient Greek etyma and reconciled these with LSJ using the skos:relatedMatch property.
The etymology of abacus is expressed with the CIDOC Conceptual Reference Model (CRM) class E89 Propositional Object11 as a borrowing by way of the lemonEty:etyLinkType property. This set-up is also valid for calques, should these become available in future.
The CRMinf extension of CRM and the Open
Vocabulary
11http://www.cidoc-crm.org/Entity/
e89-propositional-object/version-6.0
resent uncertainty as a “belief” or confidence
value
Listing 4: Uncertainty
Additionally, we employed the Dublin CoreTM
Metadata Terms vocabulary to supply the resource
with descriptive metadata, such as publisher and
licence
All editorial notes in IGVLL were excluded from the data model.
As previously mentioned, with this development LiLa’s etymological purview now covers both direct inheritance and borrowing. Figure 4, for example, shows all etymological information in the Knowledge Base associated with LiLa’s common noun muscus “moss, musk” (top row, centre node). LiLa’s “muscus” is connected to the “muscus” lexical entries of both IGVLL and the Brill Etymological Dictionary via the bidirectional OntoLex property canonicalForm. These lexical entries point to their respective etyma via the directed lemonETY etymology and etymon properties.
This paper describes the preparation and
integration of Saalfeld’s Index Graecorum Vocabulorum
in Linguam Latinam (1874) in the LiLa
Knowledge Base of Linguistic Resources for Latin. This
first list of 1; 763 Latin loans from Ancient Greek
adds 68 new Latin lemmas to LiLa, stretches its
data model to include borrowing and has been
mapped to the digitised Greek-English
LiddellScott-Jones lexicon. Beyond LiLa, this linguistic
resource might be integrated in other resources,
such as dictionaries
The data and code for the project are available at: https://github.com/CIRCSE/ index-graecorum-vocabulorum.
This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme - Grant Agreement No 769994. James Noel Adams. 2003. Bilingualism and the Latin
Language. Cambridge University Press.