Lemmatization and POS-tagging process by using joint learning approach. Experimental results on Classical Armenian, Old Georgian, and Syriac

What it is about

The authors lemmatize and POS-tag three under-resourced ancient languages (Classical Armenian, Old Georgian and Syriac) from corpora of the GREgORI project. These texts were first analysed automatically out of context with dictionaries and rules, then proofread by hand. The Syriac set (D-SYC) has 46,859 tokens, 10,612 of them unique, split 80/10/10 by sentence. They use PIE, a bidirectional-RNN lemmatizer with a character-level attention encoder-decoder that sees the sentence, with default hyperparameters, and compare a CRF decoder (LEMMING) with a linear decoder for POS tagging; the linear decoder wins on most sets. The mean accuracy is 91.63% for lemmatization and 92.56% for POS tagging. The authors report that about half of the lemmatization mistakes fall on polylexical forms (words with attached prefixes or suffixes), and that about 20% of Syriac training tokens are unknown.

Ideas

• I1. Report accuracy in three rows, all tokens, ambiguous tokens and unknown tokens, as the paper does. For tools/hbo, run the same three rows on the pointed and the unpointed test sets, so that a good average does not hide poor results on unseen words.
• I2. Count errors per component of a clitic-bearing form: split a polylexical Syriac form into prefix, stem and suffix, each with its own lemma, POS and morphology (the paper's @-separated columns), and score each part. Use this as the error unit for tools/hbo segmentation output.
• I3. Keep a small list of homographic lemmas (the paper's example is mayr, "mother" and "cedrus") and check which of them the rule-based analysis confuses. This is a test list, not a model.

For Mova

I1 and I3 are test sets and reports for tools/hbo. I2 goes into Particles and syntax of Biblical Aramaic as the scoring rule for attached prepositions and pronouns. PIE is a neural model: it may be used only as a teacher or as a comparison in measurements, never as a parser at test time.

Topics: Part-of-speech tagging

Cite

The paper itself, not this card. BibTeX from the library's record of the work.

@inproceedings{vidalgorene2020lemmatization,
  title = {{Lemmatization and POS-tagging process by using joint learning approach. Experimental results on Classical Armenian, Old Georgian, and Syriac}},
  author = {Vidal-Gorène, Chahan and Kindt, Bastien},
  booktitle = {Proceedings of LT4HALA 2020 - 1st Workshop on Language Technologies for Historical and Ancient Languages},
  year = {2020},
  url = {https://aclanthology.org/2020.lt4hala-1.4/}
}

References

45 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Arzhanov, Y. (2018). Syriac Sayings of Greek Philosophers: A Study in Syriac Gnomologia with Edition and Translation. Corpus Scriptorum Christianorum Orientalium, 669. Subsidia, 138. Peeters, Leuven. ↗
2. Boisson, P. (2014). Précis de géographie du monde indien à l’usage des commerçants: édition et traduction annotée. In A. Mardirossian, et al., editors, Mélanges Jean-Pierre Mahé, Travaux et Mémoires, 18, pages 105–126. Association des Amis du Centre d’Histoire et Civilisation de Byzance, Paris. find
3. Calzolari, V. (2017). Acta Pauli et Theclae, Prodigia Theclae, Martyrium Pauli. Corpus Christianorum. Series Apocryphorum, 20. Apocrypha Armeniaca, 1. Brepols, Turnhout. find
4. Coulie, B. and Métrévéli, H. (2001). Sancti Gregorii Nazianzeni Opera. Versio Iberica. III. Oratio XXXVIII. Corpus Christianorum. Series Graeca, 45. Corpus Nazianzenum, 12. Brepols, Turnhout. find
5. Coulie, B. and Métrévéli, H. (2004). Sancti Gregorii Nazianzeni Opera. Versio Iberica. IV. Oratio XLIII. Corpus Christianorum. Series Graeca, 52. Corpus Nazianzenum, 17. Brepols, Turnhout. find
6. Coulie, B. and Métrévéli, H. (2007). Sancti Gregorii Nazianzeni Opera. Versio Iberica. V. Orationes XXXIX et XL. Corpus Christianorum. Series Graeca, 58. Corpus Nazianzenum, 20. Brepols, Turnhout. find
7. Coulie, B. and Métrévéli, H. (2011). Sancti Gregorii Nazianzeni Opera. Versio Iberica. VI. Orationes XI, XXI, XLII. Corpus Christianorum. Series Graeca, 78. Corpus Nazianzenum, 26. Brepols, Turnhout. find
8. Coulie, B. and Sirinian, A. (1999). Sancti Gregorii Nazianzeni Opera. Versio armeniaca. III. Orationes XXI, VIII. Oratio VII. Corpus Christianorum. Series Graeca, 38. Corpus Nazianzenum, 7. Brepols, Turnhout. find
9. Coulie, B., Kindt, B., and Pataridze, T. (2013). Lemmatisation automatique des sources en géorgien ancien. Le Muséon, 126:161–201. find
10. Coulie, B., Kindt, B., and Kepeklian, G. (2020). Un jeu d’étiquettes morphosyntaxiques pour le traitement automatique de l’arménien ancien. Études Arméniennes Contemporaines. in press. find
11. Coulie, B. (1994). Sancti Gregorii Nazianzeni Opera. Versio armeniaca. I. Orationes II, XII, IX. Corpus Christianorum. Series Graeca, 28. Corpus Nazianzenum, 3. Brepols, Turnhout. find
12. Coulie, B. (1996). La lemmatisation des textes grecs et byzantins : une approche particulière de la langue et des auteurs. Byzantion, 66:35–54. find
13. Dereza, O., (2018). Lemmatization for Ancient Languages: Rules or Neural Networks?: 7th International Conference, AINL 2018, St. Petersburg, Russia, October 17–19, 2018, Proceedings, pages 35–47. Springer, Jan. find
14. Egen, S., Gleim, R., and Mehler, A. (2016). Lemmatization and morphological tagging in German and Latin: A comparison and a survey of the state-of-the-art. In Proceedings of the Tenth International Conference on Language Resources and Evaluation. find
15. Furlani, G. (1933). Le Categorie e gli Ermeneutici di Aristotele nella versione siriaca di Giorgio delle Nazioni. Serie VI. Vol. V. Fasc. 1. Reale Accademia Nazionale dei Lincei, Rome. find
16. Haelewyck, J.-C., Kindt, B., Schmidt, A., and Atas, N. (2018). La concordance bilingue grecque - syriaque des Discours de Grégoire de Nazianze. Babelao, 7:51–80. ↗
17. Haelewyck, J.-C. (2011). Sancti Gregorii Nazianzeni Opera. Versio Syriaca V. Orationes I, II, III. Corpus Christianorum. Series Graeca, 77. Corpus Nazianzenum, 25. Brepols, Turnhout. find
18. Haelewyck, J.-C. (2014). Historia Zosimi De Vita Beatorum Rechabitarum ; Édition de la version syriaque brève. Le Muséon, 127:95–147. find
19. Haelewyck, J.-C. (2015). La version syriaque longue de l’Historia Zosimi De Vita Beatorum Rechabitarum ; Édition et traduction. Le Muséon, 128:295–379. find
20. Haelewyck, J.-C. (2016). Histoire de Zosime sur la vie des Bienheureux Réchabites. Les versions orientales et leurs manuscrits. Corpus Scriptorum Christianorum Orientalium, 664. Subsidia, 135. Peeters, Leuven. find
21. Haelewyck, J.-C. (2017a). Histoire de Zosime sur la Vie des Bienheureux Réchabites. Les trois recensions syriaques. Édition de la version résumée. Parole de l’Orient, 43:175–194. find
22. Haelewyck, J.-C. (2017b). Les versions syriaques des Discours de Grégoire de Nazianze: un processus continu de révision. Babelao, 6:75–124. ↗
23. Kindt, B. (2004). La lemmatisation des sources patristiques et byzantines au service d’une description lexicale du grec ancien. Byzantion, 74:213–272. find
24. Kindt, B. (2018). Processing Tools for Greek and Other Languages of the Christian Middle East. Journal of Data Mining & Digital Humanities, Special Issue on Computer-Aided Processing of Intertextuality in Ancient Languages. ↗
25. Kondratyuk, D., Gavenčiak, T., Straka, M., and Hajič, J. (2018). LemmaTag: Jointly tagging and lemmatizing for morphologically rich languages with BRNNs. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 4921–4928, Brussels, Belgium, October-November. ACL. ↗
26. Lafontaine, G. and Coulie, B. (1983). La version arméni-enne des discours de Grégoire de Nazianze. Corpus Scriptorum Christianorum Orientalium, 446. Subsidia, 67. Peeters, Leuven. find
27. Manjavacas, E., Kádár, Á., and Kestemont, M. (2019). Improving lemmatization of non-standard languages with joint learning. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), pages 1493–1503, Minneapolis, Minnesota. Association for Computational Linguistics. ↗
28. Métrévéli, H. (1998). Sancti Gregorii Nazianzeni Opera. Versio Iberica. I. Orationes I, XLV, XLIV, XLI. Corpus Christianorum. Series Graeca, 36. Corpus Nazianzenum, 5. Brepols, Turnhout. find
29. Métrévéli, H. (2000). Sancti Gregorii Nazianzeni Opera. Versio Iberica. II. Orationes XV, XXIV, XIX. Corpus Christianorum. Series Graeca, 42. Corpus Nazianzenum, 9. Brepols, Turnhout. find
30. Müller, T., Cotterell, R., Fraser, A., and Schütze, H. (2015). Joint lemmatization and morphological tagging with lemming. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 2268–2274, Lisbon, Portugal. Association for Computational Linguistics. ↗
31. Ouzounian, A., Goepp, M., and Mutafian, C. (2012). L’inscription du régent Constantin de Papeṙōn (1241). redécouverte, relecture, remise en contexte historique. Revue des Études Arméniennes, 34:243–287. find
32. Peshiṭta Institute. (1977). The Old Testament in Syriac according to the Peshiṭta Version. Peshitta. The Old Testament in Syriac. Brill, Leiden. find
33. Sanspeur, C. (2007). Sancti Gregorii Nazianzeni Opera. Versio Armeniaca. IV. Oratio VI. Corpus Christianorum. Series Graeca, 61. Corpus Nazianzenum, 21. Brepols, Turnhout. find
34. Schmidt, A. (2002). Sancti Gregorii Nazianzeni Opera. Versio Syriaca. II. Orationes XIII, XLI. Corpus Christianorum. Series Graeca, 47. Corpus Naizanzenum, 15. Brepols, Turnhout. find
35. Sembiante, A. (2017). Appunti sulla tradizione siriaca delle opere di Gregorio Nazianzeno. Koinonia, 10:607– 634. find
36. Sirinian, A. (1999). Sancti Gregorii Nazianzeni Opera. Versio armeniaca. II. Orationes IV et V. Corpus Christianorum. Series Graeca, 37. Corpus Nazianzenum, 6. Brepols, Turnhout. find
37. Van Elverdinghe, E. (2018). Recurrent Pattern Modelling in a Corpus of Armenian Manuscript Colophons. Journal of Data Mining & Digital Humanities, Special Issue on Computer-Aided Processing of Intertextuality in Ancient Languages. ↗
38. Venturini, G. (2019). La versione siriaca della vita di Giovanni il Misericordioso di Leonzio di Neapolis. Corpus Scriptorum Christianorum, 679. Scriptores Syri, 263. Peeters, Leuven. ↗
39. Vidal-Gorène, C. and Decours-Perez, A. (2020). Languages resources for poorly endowed languages : The case study of Classical Armenian. In Proceedings of the Twelfth International Conference on Language Resources and Evaluation (LREC 2020). in press. find
40. Vidal-Gorène, C., Decours-Perez, A., Queuche, B., Ouzounian, A., and Riccioli, T. (2019). Digitalization and Enrichment of the Nor Baṙgirk‘ Haykazean Lezui: Work in Progress for Armenian Lexicography. Journal of the Society of Armenian Studies, 27. in press. 7. Language Resource References American University of Armenia. (1999). Digital Library of Armenian Literature. find
41. Calfa. (2014). Calfa - Enriched Dictionaries of Classical and Modern Armenian. find
42. J. Gippert. (2003). TITUS Project. Johann Wolfgang Goethe University. find
43. Ilia State University. (2009). Georgian Language Corpus. find
44. Université catholique de Louvain. (1990). GREgORI Project - Softwares, linguistic data and tagged corpus for ancient GREek and ORIental languages. find
45. J. E. Walters. (2004). Digital Syriac Corpus. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2020-gorene-lemmatization-pos-tagging-process-joint/