Summary

This Universal Dependencies version of the LLCT (Late Latin Charter Treebank) consists of an automated conversion of the LLCT2 treebank from the Latin Dependency Treebank (LDT) format into the Universal Dependencies standard.

Introduction

The Universal Dependencies version of the LLCT (Late Latin Charter Treebank) results from an automated conversion of the LLCT2 treebank from the Latin Dependency Treebank (LDT) format into the Universal Dependencies standard. The LLCT2 is the second part of three LLCT treebanks, the first part (LLCT1) being available in LDT format (see Korkiakangas, in press) and the third part still under construction (as of 4/2020).

The LLCT2 contains 521 Early Medieval Latin original documents (charters) written in Tuscia (Tuscany), Italy, between AD 774 and 897. They all represent the legal (documentary) genre. Their language is a non-standard variety of Latin which differs from Classical as well as from Medieval Latin in terms of spelling, morphology, and syntax. For a general overview of the LLCT treebanks, see Korkiakangas (in press).

The original annotation follows mainly the Guidelines for the Syntactic Annotation of Latin Treebanks (v. 1.3) (Bamman et al., 2007). However, the non-standard features are annotated following an additional set of rules described in (Korkiakangas and Passarotti, 2011) and in (Cecchini et al., 2020).

Sources, annotation & conversion process

The original data has been annotated automatically according to the LDT style using the first part of the LLCT treebank, i.e. the LLCT1, as the training set, and has then been manually corrected. The UD conversion has been performed automatically, based on a PERL script previously used for the Index Thomisticus treebank (https://github.com/UniversalDependencies/UD_Latin-ITTB), which has been partially rewritten (see Cecchini et al., 2018, 2020). Some minor a posteriori corrections have been applied to comply to the UD validation script.

The underlying text derives from four 19th-century editions:

1. Memorie e documenti per servire all'istoria del Ducato di Lucca 5:2. Ed. by D. Barsocchini. Lucca: Francesco Bertini, 1837;
2. Memorie e documenti per servire all'istoria del Ducato di Lucca 5:3. Ed. by D. Barsocchini. Lucca: Francesco Bertini, 1841;
3. Memorie e documenti per servire all'istoria del Ducato di Lucca 4:1. Ed. by D. Bertini. Lucca: Francesco Bertini, 1818;
4. Memorie e documenti per servire all'istoria del Ducato di Lucca 4:2. Ed. by D. Bertini. Lucca: Francesco Bertini, 1836.

Their text has been manually corrected to meet modern editorial standards (see Korkiakangas, in press). In the LLCT2, each sentence receives a unique identifier that refers to its numbering in the modern volumes of the Chartae Latinae Antiquiores (Cavallo et al., 1997-2019). For example, the identifier document_id='36:1047'-span='6' means that the sentence at hand can be found in charter no. 1047 of volume no. 36, and is the 6th sentence thereof (in the LLCT2, sentences are bound by the punctuation marks ., ; and :).

Splitting of the set

The corpus has been split into three sets for training (train), development (dev) and testing (test), roughly following an 80%/10%/10% ratio. We notice that, in the current UD release, 223 sentences have been left out from the original LLCT2 corpus, as their conversion requires further consideration regarding how they have to be restructured into the UD standard. This means that some of the charters might display a gap (e.g., charter 36:1049 goes from sentence 30 to 32). All sentences are in the order in which they appear in their respective charters.

A total of 242411 tokens is thus subdivided into three sets of 194143 (train, 7289 sentences), 24189 (dev, 850) and 24079 (test, 884) tokens each, and care has been taken not to break up the same charter over different sets.

Acknowledgements

The conversion of the LLCT2 into the Universal Dependencies was realized by Flavio Massimiliano Cecchini and Marco Passarotti within the LiLa project (Linking Latin, https://lila-erc.eu/). The original LLCT2 was annotated by Timo Korkiakangas under projects funded by the University of Oslo and the Academy of Finland.

References

• Bamman, D., Crane, G., Passarotti, M., and Raynaud, S. (2007). Guidelines for the Syntactic Annotation of Latin Treebanks (v. 1.3). Tufts Published Scholarship. Tufts University's Digital Collections and Archives, Medford, MA, USA. Permanent URL: http://hdl.handle.net/10427/42683 or https://itreebank.marginalia.it/doc/2007_Passa+Bamman+Crane+Raynaud_Guidelines%20Tb.pdf

• Cavallo, G., Nicolaj, G., et al., editors (1997-2019). Chartae Latinae Antiquiores: facsimile edition of the Latin charters, 2nd series: ninth century. Urs Graf Verlag, Dietikon, Zürich, Switzerland.

• Cecchini, F. M., Korkiakangas, T. and Passarotti, M. (2020). A New Latin Treebank for Universal Dependencies: Charters between Ancient Latin and Romance Languages. In Proceedings of the Twelfth International Conference on Language Resources and Evaluation (LREC2020), Marseille, France, May. European Language Resources Association (ELRA).

• Cecchini, F. M., Passarotti, M., Marongiu, P., and Zeman, D. (2018). Challenges in Converting the Index Thomisticus Treebank into Universal Dependencies. In Proceedings of the Second Workshop on Universal Dependencies (UDW 2018) at EMNLP 2018, pages 27–36, Bruxelles, Belgium. Special Interest Group on linguistic DATa and corpus-based approaches to NLP (SIGDAT), ACL. Available at https://www.aclweb.org/anthology/W18-6004.pdf

• Gamba, F. and Zeman, D. (2023a). Universalising Latin Universal Dependencies: a harmonisation of Latin treebanks in UD. In Proceedings of the Sixth Workshop on Universal Dependencies (UDW, GURT/SyntaxFest 2023), Washington, DC, USA, March. Association for Computational Linguistics (ACL).

• Gamba, F. and Zeman, D. (2023b). Latin Morphology through the Centuries: Ensuring Consistency for Better Language Processing. In Proceedings of the Ancient Language Processing Workshop associated with the 14th International Conference on Recent Advances in Natural Language Processing RANLP 2023, Varna, Bulgaria, September.

• Korkiakangas, T. (in press). Late Latin Charter Treebank: contents and annotation, in Corpora, 16 (2021). Preprint: https://researchportal.helsinki.fi/en/publications/late-latin-charter-treebank-contents-and-annotation

• Korkiakangas, T. and Passarotti, M. (2011). Challenges in Annotating Medieval Latin Charters. Journal for Language Technology and Computational Linguistics (JLCL), 26(2):105–116, November. In Francesco Mambrini, et al., editors, Annotation of Corpora for Research in the Humanities: Proceedings of the ACRH Workshop, Heidelberg, 5. Jan. 2012.

Changelog

• 2023-11-15 v2.13
  • Implementation of overall syntactic and morphologic harmonisation following (Gamba & Zeman 2023a) and (Gamba & Zeman 2023b)
  • Orthographic univerbation of the phrase ne c by means of multi-word tokens, retaining the individual analyses of its components
  • Complete re-annotation of 16 gapping (i.e. non-projective) fixed constructions (others have been kept for the moment since they are not trivially resolved)
• 2022-11-15 v2.11
  • Implementations of the amendment for clausal non-verbal copular constructions and corrections regarding multiple subjects and introduction of the :outer subtype for subjects and copulas
  • Implementation of the proposal of VerbForm reform as for (Cecchini, 2021; see documentation entry for VerbForm)
    • at the same time, introduction of the TraditionalMood and TraditionalTense features in MISC to take into account traditional denominations
  • Double-pronoun constructions, also called free relatives, have been highlighted semi-automatically by means of the the newly introduced :relcl subtype
  • Implementation of the treatment of personal pronouns as Person-bearing determiners (relation det, part of speech PRON) of (proper) nouns, instead as of heads modified by appositions (relation appos), as discussed in (Höhn, 2021; UDW 2021, Sofia)
  • Normalisation of textual gap annotation: no more pseudolemma missing^token, but empty lemma (_), empty part of speech (X) if not retrievable from the context, and possibly empty relation (dep) and convenience head attachment if not retrievable from the context
  • Some instances of the correlating construction tam ... quam et 'so ... as also' have been (partially semi-automatically) normalised from co-ordinating conjunctions to focaliser + comparative adverbial clause
  • Many "small words" (especially particles) and their features have started being standardised among treebanks. For example, is/ea/id is now always a PRON with PronType=Prs, nam always a PART, AdvType has been added to many "adverbs"...
    • et, nec etc. now stay CCONJ even when acting as focalisers
    • unus is uniformed to a determiner (DET) with both cardinal numeral type and indefinite pronoun type, and numeric value 1
    • multiple (comma-separated) PronTypes have been revised and reduced to one; the interrogative Int value has been streamlined as possible
  • Remotion of the value Pos for the feature Degree ("positive" degree is still retrievable as the absence of another degree)
  • Various other occasional corrections of the annotation
    • The LLCT is still waiting for the introduction of inflectional classes and some other lexical features, a complex made more difficult by the particular Latin of these texts
• 2020-05-15 v2.6
  • Initial release in Universal Dependencies.

=== Machine-readable metadata (DO NOT REMOVE!) ================================
Data available since: UD v2.6
License: CC BY-SA 4.0
Includes text: yes
Genre: nonfiction legal
Lemmas: converted from manual
UPOS: converted from manual
XPOS: manual native
Features: converted from manual
Relations: converted with corrections
Contributors: Korkiakangas, Timo; Cecchini, Flavio Massimiliano; Passarotti, Marco
Contributing: elsewhere
Contact: flavio.cecchini@unicatt.it, timo.t.korkiakangas@gmail.com, marco.passarotti@unicatt.it
===============================================================================