A Rich Morphological Tagger for English: Exploring the Cross-Linguistic Tradeoff Between Morphology and Syntax

What it is about

The question is whether English, a morphologically poor language, carries in its syntax the same meanings that Czech marks with inflection. The authors project rich morphological tags (UniMorph: case, number, person, tense, voice, and more) from Czech words onto their aligned English words in the Prague Czech–English treebank, using GIZA++ alignments and a rule for many-to-many links. Before training, they check the projections against PTB tags where both should agree (for example VBZ with third-person singular). A feed-forward network with two hidden layers and features from automatic dependency and CFG parses (head word and head POS, head chains, child determiners and case markers, distance to root) is trained on WSJ 02–21 and tested on WSJ 22. Whole-tag accuracy is 93.7%, against 14.1% for a most-frequent-value baseline. Case is the hardest subtag (69.0%, against 35.0% for the baseline), and dependency features help most for case.

Ideas

I1. en/src/udreg.rs and en/src/morph.rs: a rule-based Case feature read from the arcs. Give a noun Acc when it is a obj or iobj of a verb, Nom when it is nsubj, and the case of its case child otherwise. Report accuracy per class on UD EWT dev, against the majority baseline the paper uses.
I2. Use the paper's split of features as a test design: report each subtag separately (case, number, tense, person) and never only the whole-tag score, so that a gain in one category is not hidden.
I3. Negative control: shuffle the head labels within each sentence and re-run the same rule. The score must fall, or the rule is not reading the tree.
I4. Projection check: for any cross-language projection Mova builds (for example from the Early Modern texts), check the projected value against the expected English form on words where both must agree, as the paper does with Table 2.

For Mova

The chapters are 11. Nouns (case, number) and 13. Verbs (tense, person). The code is en/src/udreg.rs (the feature registry), en/src/morph.rs, and the parse output of en/src/parse.rs. The measure is per-feature accuracy on UD EWT dev, run twice to check determinism. The neural tagger is not for Mova; the feature design and the per-category baseline are.

Topics: Part-of-speech tagging

Cite

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

@inproceedings{kirov2017rich,
  title = {{A Rich Morphological Tagger for English: Exploring the Cross-Linguistic Tradeoff Between Morphology and Syntax}},
  author = {Kirov, Christo and Sylak-Glassman, John and Knowles, Rebecca and Cotterell, Ryan and Post, Matt},
  booktitle = {Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers},
  year = {2017},
  url = {https://aclanthology.org/E17-2018/}
}

References

17 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

Heidi Harley. 2015. The syntax-morphology interface. In Tibor Kiss and Artemis Alexiadou, editors, Syntax - Theory and Analysis: An International Handbook, volume II, pages 1128–1153. Mouton de Gruyter, Berlin. find
Martı́n Abadi, Ashish Agarwal, Paul Barham, Eugene Brevdo, Zhifeng Chen, Craig Citro, Greg S. Corrado, Andy Davis, Jeffrey Dean, Matthieu Devin, Sanjay Ghemawat, Ian Goodfellow, Andrew Harp, Geoffrey Irving, Michael Isard, Yangqing Jia, Rafal Jozefowicz, Lukasz Kaiser, Manjunath Kudlur, Josh Levenberg, Dan Mané, Rajat Monga, Sherry Moore, Derek Murray, Chris Olah, Mike Schuster, Jonathon Shlens, Benoit Steiner, Ilya Sutskever, Kunal Talwar, Paul Tucker, Vincent Vanhoucke, Vijay Vasudevan, Fernanda Viégas, Oriol Vinyals, Pete Warden, Martin Wattenberg, Martin Wicke, Yuan Yu, and Xiaoqiang Zheng. 2015. TensorFlow: Large-scale machine learning on heterogeneous systems. Software available from tensorflow.org. find
Mitchell P. Marcus, Mary Ann Marcinkiewicz, and Beatrice Santorini. 1993. Building a large annotated corpus of English: The Penn Treebank. Computational Linguistics, 19(2):313–330. find our card
Thomas Müller, Ryan Cotterell, Alexander Fraser, and Hinrich Schütze. 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, September. Association for Computational Linguistics. find
Eleftherios Avramidis and Philipp Koehn. 2008. Enriching morphologically poor languages for statistical machine translation. In Proceedings of ACL-08: HLT, pages 763–770, Columbus, Ohio, June. Association for Computational Linguistics. find
Franz Josef Och and Hermann Ney. 2003. A systematic comparison of various statistical alignment models. Computational Linguistics, 29(1):19–51, March. find
Jeffrey Pennington, Richard Socher, and Christopher D. Manning. 2014. Glove: Global vectors for word representation. In Empirical Methods in Natural Language Processing (EMNLP), pages 1532– 1543. find
Jan Buys and Jan A. Botha. 2016. Cross-lingual morphological tagging for low-resource languages. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (ACL), pages 1954–1964, Berlin, August. Association for Computational Linguistics. find
Rico Sennrich and Barry Haddow. 2016. Linguistic input features improve neural machine translation. In Proceedings of the 1st Conference on Machine Translation, volume 1, pages 83–91, Berlin, August. Association for Computational Linguistics. find
Elliott Franco Drábek and David Yarowsky. 2005. Induction of fine-grained part-of-speech taggers via classifier combination and crosslingual projection. In Proceedings of the ACL Workshop on Building and Using Parallel Texts, pages 49–56, Ann Arbor, June. Association for Computational Linguistics. find
John Sylak-Glassman, Christo Kirov, David Yarowsky, and Roger Que. 2015. A language-independent feature schema for inflectional morphology. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2: Short Papers), pages 674– 680, Beijing, China, July. Association for Computational Linguistics. find
Rong-En Fan, Kai-Wei Chang, Cho-Jui Hsieh, Xiang-Rui Wang, and Chih-Jen Lin. 2008. LIBLINEAR: A library for large linear classification. Journal of Machine Learning Research, 9:1871–1874. find
Xavier Glorot, Antoine Bordes, and Yoshua Bengio. 2011. Deep sparse rectifier neural networks. In International Conference on Artificial Intelligence and Statistics, pages 315–323. find
Ke Tran, Arianna Bisazza, and Christof Monz. 2015. A distributed inflection model for translating into morphologically rich languages. In Proceedings of MT-Summit 2015. find
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, İstanbul, Turkey. ELRA, European Language Resources Association. find
David Yarowsky, Grace Ngai, and Richard Wicentowski. 2001. Inducing multilingual text analysis tools via robust projection across aligned corpora. In HLT ’01 Proceedings of the First International Conference on Human Language Technology Research, pages 1–8, Stroudsburg, PA. Association for Computational Linguistics. find
Jan Hajič. 2000. Morphological tagging: Data vs. dictionaries. In Proceedings of the 1st North American chapter of the Association for Computational Linguistics Conference (NAACL 2000), pages 94–101, Seattle, May. Association for Computational Linguistics. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/kirov-2017-rich-morphological-tagger-english-exploring/