Translation of "It" in a Deep Syntax Framework

What it is about

English "it" has no single Czech counterpart. It may be a personal pronoun or zero anaphor, a neuter demonstrative "to", or nothing at all in a cleft or pleonastic construction. The paper builds a discriminative classifier for these three classes inside the TectoMT deep-syntax system. The training data is the Prague Czech-English Dependency Treebank, where the class of each "it" is taken from the Czech subtree aligned to the English parent verb, with the same functor. Only about a sixth of the training instances were labelled by hand. The features come from the parse, from an anaphoricity resolver, from co-occurrence counts of functor and parent lemma with the Czech class over CzEng 1.0 (binned on a log scale), and from gold coreference for one setting. On a 543-instance test set, the best classifier reaches about 76% accuracy, against 59% for the most frequent class. The BLEU score does not move (0.1404 to 0.1403), but a manual check of 50 changed sentences favours the new model in 24 cases and the old in 13.

Ideas

• I1. A pronoun class on meta tokens: give each "it" (and the Bible's "it", "thou", "ye") a class of personal, demonstrative or null, from the same parse features, and write the class into the meta token (open problem 4). The encoder then chooses the surface form from the class.
• I2. Get the class from the parent, not the pronoun: the paper takes the class from the aligned parent verb, because pronoun alignment is unreliable. Use the same trick in tools/translate/src/hebrew.rs for Hebrew pronominal suffixes, where IBM Model 1 aligns poorly.
• I3. Count-based lexical features from a large parallel corpus, binned by log count, as the paper does with CzEng: use the English-to-English KJV/WEB pairs in the translator, and keep the bin edges fixed in the code so the run is deterministic.
• I4. Report the manual check: a sentence-level comparison of old and new output, with counts of better, worse and equal, as the paper does. Use this on the author deltas (open problem 8), not only on BLEU or chrF.

For Mova

The code is tools/translate/src/meta.rs (the classes) and tools/translate/src/profile.rs (the choice of surface form). The classifier here is a small discriminative model, not a neural one, so the approach carries over directly. Its main limit is the small hand-labelled set; Mova would need to measure how many labels each class needs before the gain shows.

Topics: Machine translation

Cite

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

@inproceedings{novak2013translation,
  title = {{Translation of "It" in a Deep Syntax Framework}},
  author = {Novák, Michal and Nedoluzhko, Anna and Žabokrtský, Zdeněk},
  booktitle = {Proceedings of the Workshop on Discourse in Machine Translation},
  year = {2013},
  url = {https://aclanthology.org/W13-3307/}
}

References

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

1. Mohit Bansal and Dan Klein. 2012. Coreference Semantics from Web Features. In Proceedings of the 50th Annual Meeting of the ACL: Long Papers – Volume 1, pages 389–398, Stroudsburg, PA, USA. Association for Computational Linguistics. find
2. Shane Bergsma and David Yarowsky. 2011. NADA: A Robust System for Non-Referential Pronoun Detection. In DAARC, pages 12–23, Faro, Portugal, October. find
3. Ondřej Bojar, Zdeněk Žabokrtský, Ondřej Dušek, Petra Galuščáková, Martin Majliš, David Mareček, Jiřı́ Maršı́k, Michal Novák, Martin Popel, and Aleš Tamchyna. 2012. The Joy of Parallelism with CzEng 1.0. In Proceedings of LREC 2012, Istanbul, Turkey, May. ELRA, European Language Resources Association. find
4. Ondřej Bojar, Rudolf Rosa, and Aleš Tamchyna. 2013. Chimera – Three Heads for English-to-Czech Translation. In Proceedings of the Eight Workshop on Statistical Machine Translation. Under review. find
5. Chris Callison-Burch, Philipp Koehn, Christof Monz, and Omar Zaidan. 2011. Findings of the 2011 Workshop on Statistical Machine Translation. In Proceedings of the Sixth Workshop on Statistical Machine Translation, pages 22–64, Edinburgh, Scotland, July. Association for Computational Linguistics. find
6. Liane Guillou. 2012. Improving Pronoun Translation for Statistical Machine Translation. In Proceedings of the Student Research Workshop at the 13th Conference of the EACL, pages 1–10, Stroudsburg, PA, USA. Association for Computational Linguistics. find
7. Jan Hajič, Eva Hajičová, Jarmila Panevová, Petr Sgall, Silvie Cinková, Eva Fučı́ková, Marie Mikulová, Petr Pajas, Jan Popelka, Jiřı́ Semecký, Jana Šindlerová, Jan Štěpánek, Josef Toman, Zdeňka Urešová, and Zdeněk Žabokrtský. 2011. Prague Czech-English Dependency Treebank 2.0. find
8. Jan Hajič, Eva Hajičová, Jarmila Panevová, Petr Sgall, Ondřej Bojar, Silvie Cinková, Eva Fučı́ková, Marie Mikulová, Petr Pajas, Jan Popelka, Jiřı́ Semecký, Jana Šindlerová, Jan Štěpánek, Josef Toman, Zdeňka Urešová, and Zdeněk Žabokrtský. 2012. Announcing Prague Czech-English Dependency Treebank 2.0. In Proceedings of the 8th International Conference on Language Resources and Evaluation (LREC 2012), pages 3153–3160. ELRA. find
9. Christian Hardmeier and Marcello Federico. 2010. Modelling Pronominal Anaphora in Statistical Machine Translation. In Marcello Federico, Ian Lane, Michael Paul, and François Yvon, editors, Proceedings of the seventh International Workshop on Spoken Language Translation (IWSLT), pages 283–289. find
10. Philipp Koehn, Franz Josef Och, and Daniel Marcu. 2003. Statistical Phrase-based Translation. In Proceedings of the 2003 Conference of the NAACL HLT – Volume 1, pages 48–54, Stroudsburg, PA, USA. Association for Computational Linguistics. find
11. John Langford. 2012. Vowpal Wabbit. find
12. Ronan Le Nagard and Philipp Koehn. 2010. Aiding Pronoun Translation with Co-Reference Resolution. In Proceedings of the Joint Fifth Workshop on Statistical Machine Translation and MetricsMATR, pages 252–261, Uppsala, Sweden, July. Association for Computational Linguistics. find
13. Linguistic Data Consortium. 1999. Penn Treebank 3. LDC99T42. find
14. David Mareček, Zdeněk Žabokrtský, and Václav Novák. 2008. Automatic Alignment of Czech and English Deep Syntactic Dependency Trees. In Proceedings of the Twelfth EAMT Conference, pages 102–111. find
15. Fabian Pedregosa, Gaël Varoquaux, Alexandre Gramfort, Vincent Michel, Bertrand Thirion, Olivier Grisel, Mathieu Blondel, Peter Prettenhofer, Ron Weiss, Vincent Dubourg, Jake Vanderplas, Alexandre Passos, David Cournapeau, Matthieu Brucher, Matthieu Perrot, and Édouard Duchesnay. 2011. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 12:2825–2830, November. find
16. Petr Sgall, Eva Hajičová, and Jarmila Panevová. 1986. The Meaning of the Sentence in Its Semantic and Pragmatic Aspects. D. Reidel Publishing Company, Dordrecht. find
17. Petr Sgall. 1967. Generativnı́ popis jazyka a česká deklinace. Academia, Prague, Czech Republic. find
18. Kateřina Veselovská, Giang Linh Nguy, and Michal Novák. 2012. Using Czech-English Parallel Corpora in Automatic Identification of It. In The Fifth Workshop on Building and Using Comparable Corpora, pages 112–120. find
19. Zdeněk Žabokrtský, Jan Ptáček, and Petr Pajas. 2008. TectoMT: Highly Modular MT System with Tectogrammatics Used as Transfer Layer. In Proceedings of the Third Workshop on Statistical Machine Translation, pages 167–170, Stroudsburg, PA, USA. Association for Computational Linguistics. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/novak-2013-translation-it-deep-syntax-framework/