Diachronic proximity vs. data sparsity in cross-lingual parser projection. A case study on Germanic

What it is about

The question is whether a dependency parser for a historical Germanic variety with little training data can be helped by data from related varieties. Modern English parse trees are projected onto parallel Bible verses in Old English, Middle English, Early Modern High German and Middle Icelandic, using word alignment. Each target word gets a hyperlemma (its aligned English word). A fragment-aware parser is trained on 437 projected sentences per language and tested on 174. The baseline UAS, from target data alone, is .60 for Middle English, .31 for Old English, .41 for German and .32 for Icelandic. Adding data from related varieties helps only German, with a significant effect; Icelandic gets significantly worse when Western Germanic data is added. The authors suggest that varieties separated by about 750 years or less can help each other, while those separated by more than 1,000 years cannot.

Ideas

• I1. 12. How Mova reads this English today: pipeline, fixes, failures: project the Mova English parse onto aligned KJV and Wycliffe verses, then train the perceptron on a Middle English set plus the Early Modern set and measure UAS on a held-out gospel. Run twice, and report the per-pair gain with a significance test, as the paper does.
• I2. Use the closed table in en/src/emode.rs as the hyperlemma: map thee, hath, shalt to their modern lemmas before parsing, and measure whether the parser gains on archaic verses.
• I3. Negative control: shuffle the alignment between source and target verses. A real projection effect must disappear under this shuffle.

For Mova

The chapters are 1. The period, the register and the texts Mova reads (which periods and texts count as related) and 9. The subjunctive, conditions and wishes and 08-word-order.md (where EModE syntax departs from modern English). The code is en/src/parse.rs, en/src/emode.rs and tools/gutenberg. The measure is UAS on a held-out gospel, against the target-only baseline, with the shuffled-alignment control. Old English is not yet a Mova chapter; the paper's table is a reason to add it.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.3115/v1/w14-5302

@inproceedings{sukhareva2014diachronic,
  title = {{Diachronic proximity vs. data sparsity in cross-lingual parser projection. A case study on Germanic}},
  author = {Sukhareva, Maria and Chiarcos, Christian},
  booktitle = {Proceedings of the First Workshop on Applying NLP Tools to Similar Languages, Varieties and Dialects},
  year = {2014},
  doi = {10.3115/v1/w14-5302},
  url = {https://doi.org/10.3115/v1/w14-5302}
}

References

27 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Miguel Ballesteros and Joakim Nivre. 2012. Maltoptimizer: A system for maltparser optimization. In Nicoletta Calzolari (Conference Chair), Khalid Choukri, Thierry Declerck, Mehmet Uur Doan, Bente Maegaard, Joseph Mariani, Asuncion Moreno, Jan Odijk, and Stelios Piperidis, editors, Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC’12), Istanbul, Turkey, may. European Language Resources Association (ELRA). find
2. Marcel Bollmann, Florian Petran, and Stefanie Dipper. 2011. Rule-based normalization of historical texts. In Proceedings of the Workshop on Language Technologies for Digital Humanities and Cultural Heritage (LaTeCH-2011), pages 34–42, Hissar, Bulgaria, September. find our card
3. Marcel Bollmann. 2013. POS tagging for historical texts with sparse training data. In Proceedings of the 7th Linguistic Annotation Workshop and Interoperability with Discourse, pages 11–18, Sofia, Bulgaria, August. Association for Computational Linguistics. find our card
4. Michael Collins, Lance Ramshaw, Jan Hajič, and Christoph Tillmann. 1999. A statistical parser for czech. In Proceedings of the 37th annual meeting of the Association for Computational Linguistics on Computational Linguistics, pages 505–512. Association for Computational Linguistics. find
5. Michael Cummings. 2010. An Introduction to the Grammar of Old English: A Systemic Functional Approach. Functional Linguistics. Equinox Publishing Limited. find
6. Robert P. Ebert. 1976. Infinitival complement constructions in Early New High German. Linguistische Arbeiten. De Gruyter. find
7. Richard Johansson and Pierre Nugues. 2007. Extended constituent-to-dependency conversion for English. In Proceedings of NODALIDA 2007, pages 105–112, Tartu, Estonia, May 25-26. find our card
8. Daisuke Kawahara and Kiyotaka Uchimoto. 2007. Minimally lexicalized dependency parsing. In Proceedings of the 45th Annual Meeting of the ACL on Interactive Poster and Demonstration Sessions, pages 205–208. Association for Computational Linguistics. find
9. Mike Kestemont, Walter Daelemans, and Guy De Pauw. 2010. Weigh your words: Memory-based lemma-retrieval for Middle Dutch literary texts. In CLIN 2010. Computational linguistics in the Netherlands 20, Utrecht, The Netherlands, May. find
10. Roland Meyer. 2011. New wine in old wineskins? Tagging Old Russian via annotation projection from modern translations. Russian linguistics, 35(2):267–281. find
11. Graham Neubig, Taro Watanabe, Shinsuke Mori, and Tatsuya Kawahara. 2012. Machine translation without words through substring alignment. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 165–174, Jeju Island, Korea, July. Association for Computational Linguistics. find
12. Joakim Nivre. 2003. An efficient algorithm for projective dependency parsing. In Proceedings of the 8th International Workshop on Parsing Technologies (IWPT). find our card
13. Franz Josef Och and Hermann Ney. 2003. A systematic comparison of various statistical alignment models. Computational linguistics, 29(1):19–51. find
14. Marco Pennacchiotti and Fabio Massimo Zanzotto. 2008. Natural language processing across time: An empirical investigation on italian. In Advances in Natural Language Processing, pages 371–382. Springer. find
15. Paul Rayson, Dawn Archer, Alistair Baron, Jonathan Culpeper, and Nicholas Smith. 2007. Tagging the Bard: Evaluating the accuracy of a modern POS tagger on Early Modern English corpora. In Proceedings of the 4th Corpus Linguistics Conference (CL-2007), Birmingham, UK. find
16. Philip Resnik, Mari Broman Olsen, and Mona Diab. 1997. Creating a parallel corpus from the book of 2000 tongues. In Proc. of the Text Encoding Initiative 10th Anniversary User Conference (TEI-10). find
17. Eirı́kur Rögnvaldsson, Anton Karl Ingason, Einar Freyr Sigurdhsson, and Joel Wallenberg. 2012. The Icelandic Parsed Historical Corpus (IcePaHC). In LREC, pages 1977–1984. find our card
18. Silke Scheible, Richard J. Whitt, Martin Durrell, and Paul Bennett. 2011. Evaluating an ’off-the-shelf’ POS-tagger on Early Modern German text. In Proceedings of the 5th ACL-HLT Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities (LaTeCH-2011), pages 19–23, Portland, OR, USA, June. find
19. Kathrin Spreyer, Lilja Ovrelid, and Jonas Kuhn. 2010. Training parsers on partial trees: A cross-language comparison. In Proc. of the 7th International Conference on Language Resources and Evaluation (LREC-2010), Valletta, Malta, May. find
20. Ann Taylor, Mitchell Marcus, and Beatrice Santorini. 2003a. The Penn treebank: an overview. In Treebanks, pages 5–22. Springer. find
21. Ann Taylor, Anthony Warner, Susan Pintzuk, and Frank Beths. 2003b. The york-toronto-helsinki parsed corpus of old english prose. University of York. find
22. Krisztina Tóth, Richárd Farkas, and András Kocsor. 2008. Sentence alignment of hungarian-english parallel corpora using a hybrid algorithm. Acta Cybern., 18(3):463–478, January. find
23. C. Trips. 2002. From OV to VO in Early Middle English. Linguistics today. John Benjamins Pub. find
24. A. van Kemenade and B. Los. 2009. The Handbook of the History of English. Blackwell Handbooks in Linguistics. John Wiley & Sons. find
25. Hiroyasu Yamada and Yuji Matsumoto. 2003. Statistical dependency analysis with support vector machines. In Proceedings of IWPT. Vol. 3. 2003. find our card
26. David Yarowski and Grace Ngai. 2001. Inducing multilingual POS taggers and NP bracketers via robust projection across aligned corpora. In Proceedings of NAACL 2001, pages 200–207. find
27. Daniel Zeman and Philip Resnik. 2008. Cross-language parser adaptation between related languages. In IJCNLP-08 Workshop on NLP for Less Privileged Languages, pages 35–41, Hyderabad, India, Jan. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/sukhareva-2014-diachronic-proximity-vs-data-sparsity/