Computational Methods for the Analysis of Complementizer Variability in Language and Literature: The Case of Hebrew "she-" and "ki"

What it is about

Modern Hebrew has two complementizers, še- and ki, that are usually described as interchangeable with ki as the more formal register. The authors parse a 29-million-sentence Hebrew corpus with DictaBERT-Parse and collect every verb with a ccomp dependent, taking the earliest token of the clause as the complementizer. This gives a propensity for each verb (for example, "said" is about 55% ki and 45% še-, while "told" is about 96% še-). Next, a masked BERT is run over each complement clause, and the complementizer's embedding is clustered per verb; a cluster is "divergent" when its Jensen–Shannon divergence is above 0.04. The authors publish the table for the 100 most frequent verbs. Applied to ten novels by Amos Oz, a K-nearest-neighbour classifier flags sentences where a context strongly prefers one complementizer (over 95%) and the author chose the other. Two examples are a biblical-style ki after "told" and a še- after "inform" in a letter, both read as deliberate.

Ideas

• I1. Turn the propensity table into a rule table for the Modern Hebrew syntax: for each governing verb, which complementizer is preferred and how strongly. The table is a lookup that a he parser can use to disambiguate the complementizer, and it is checkable against the 100-verb list.
• I2. A deterministic version of the divergence test: instead of BERT clusters, cluster on surface features that are known to matter, such as a following pronominal suffix or a percent sign after the complementizer (the paper's Figure 3 and example (4) show ki is forced before digits and %). Measure how many divergent clusters survive without BERT.
• I3. Mark "biblical" ki after a speech verb (example (6), Genesis 3:11 and similar) as an archaism flag in the pragmatic layer, so that a rare complementizer triggers a register note rather than an error.

For Mova

I1 goes into 7. Syntax (complementizers) and Prepositions, prefixes, conjunctions and particles. I2 is a test for the tools/langseg and he segmentation of ki before digits, and a measure for the rule-based version. I3 belongs to Pronouns, particles and prepositions (ki as complementizer in Biblical Hebrew) and 9. Pragmatics and politeness.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/2024.nlp4dh-1.29

@inproceedings{shmidman2024computational,
  title = {{Computational Methods for the Analysis of Complementizer Variability in Language and Literature: The Case of Hebrew "she-" and "ki"}},
  author = {Shmidman, Avi and Rubinstein, Aynat},
  booktitle = {Proceedings of the 4th International Conference on Natural Language Processing for Digital Humanities},
  year = {2024},
  doi = {10.18653/v1/2024.nlp4dh-1.29},
  url = {https://doi.org/10.18653/v1/2024.nlp4dh-1.29}
}

References

19 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Inge Bartning and Suzanne Schlyter. 2004. Itinéraires acquisitionnels et stades de développement en français L2. Journal of French Language Studies, 14(3):281–299. find
2. Sacha Beniamine, Martin Maiden, and Erich Round. 2020. Opening the Romance verbal inflection dataset 2.0: A CLDF lexicon. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 3027–3035, Marseille, France. European Language Resources Association. find
3. Hanna Fadida, Alon Itai, and Shuly Wintner. 2014. A Hebrew verb–complement dictionary. Language Resources & Evaluation, 48:249–278. find
4. Anastasia Giannakidou and Alda Mari. 2021. Truth and Veridicality in Grammar and Thought: Mood, Modality, and Propositional Attitudes. University of Chicago Press, Chicago. find
5. Benjamin Kane, Will Gantt, and Aaron Steven White. 2021. Intensional gaps: Relating veridicality, factivity, doxasticity, bouleticity, and neg-raising. In Proceedings of Semantics and Linguistic Theory (SALT) 31, pages 570–605. find
6. Matthew Kanwit and Kimberly L. Geeslin. 2018. Exploring lexical effects in second language interpretation: The case of mood in Spanish adverbial clauses. Studies in Second Language Acquisition, 40(3):579– 603. find
7. Geoffrey Khan, editor. 2013. Encyclopedia of Hebrew Language and Linguistics. Brill. find
8. Ron Kuzar. 1993. Nominal clauses in Israeli Hebrew. Hebrew Linguistcs, 36:71–89. In Hebrew. find
9. Danit Yshaayahu Levi and Reut Tsarfaty. 2024. A truly joint neural architecture for segmentation and parsing. find our card
10. Amos Oz. 1986. Black Box. Am Oved Publishers. find
11. Deniz Özyıldız, Ciyang Qing, Floris Roelofsen, Maribel Romero, and Wataru Uegaki. 2023. A crosslinguistic database for combinatorial and semantic properties of attitude predicates. In Proceedings of the 5th Workshop on Research in Computational Linguistic Typology and Multilingual NLP, pages 65–75, Dubrovnik, Croatia. Association for Computational Linguistics. find
12. Divna Petkovic and Victor Rabiet. 2016. La polysémie lexicale et syntaxique de l’alternance modale indicatif/subjonctif – perspectives TAL (lexical and syntactic polysemy of the modal alternation indicative/subjunctive – NLP perspectives). In Actes de la conférence conjointe JEP-TALN-RECITAL 2016. volume 3 : RECITAL, pages 80–93, Paris, France. AFCP - ATALA. find
13. Paul Portner and Aynat Rubinstein. 2020. Desire, belief, and semantic composition: variation in mood selection with desire predicates. Natural Language Semantics, pages 343–393. find
14. Shaltiel Shmidman, Avi Shmidman, Moshe Koppel, and Reut Tsarfaty. 2024. MRL parsing without tears: The case of Hebrew. In Findings of the Association for Computational Linguistics ACL 2024, pages 4537–4550, Bangkok, Thailand and virtual meeting. Association for Computational Linguistics. find our card
15. Aaron White and Kyle Rawlins. 2020. Frequency, acceptability, and selection: A case study of clause-embedding. Glossa: a journal of general linguistics, 5(1):105. find
16. Ghil‘ad Zuckermann. 2006. Complement clause types in Israeli. In Dixon R. M. W. and Aikhenvald Alexandra Y., editors, Complementation: A Cross-Linguistic Typology, Vol. 3. Oxford University Press. find
17. Alda Mari and Paul Portner. 2021. Mood variation with belief predicates: Modal comparison and the raisability of questions. Glossa: a journal of general linguistics, 40(1). find
18. Caterina Mauri and Andrea Sansò. 2016. The Linguistic Marking of (Ir)Realis and Subjunctive. In The Oxford Handbook of Modality and Mood. Oxford University Press. find
19. Ellise Moon and Aaron Steven White. 2020. The source of nonfinite temporal interpretation. In Proceedings of the 50th Annual Meeting of the North East Linguistic Society (NELS), pages 11–24. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2024-shmidman-computational-methods-analysis-complementizer-variability/