Elamur Research⌕
☰
home / papers / 2017 / Cross-Lingual Parser Selection for Low-Resource Languages

Cross-Lingual Parser Selection for Low-Resource Languages

Summary

Delexicalized parsers operate only on sequences of universal POS tags. They can therefore be used to analyze the syntactic dependencies of almost any language. The question remains which source-language model to apply to a given text. Usually the best parser is selected based on results on the target-language test data. In essence, this is an oracle that is not available in real-world conditions. The author proposes a lightweight method for selecting the source at system run time. The language of the input text is identified automatically, and its proximity to the source languages is estimated with three measures: the divergence of POS-trigram distributions (KL-POS), the probability from the langid.py language identifier (LANG-ID), and the similarity of typological features in the WALS database. These measures are combined linearly, and the parses of all parsers are merged via reparsing (maximum spanning tree). The experiments cover 26 languages from UD. Cross-lingual POS taggers are used, and the input size varies down to a single sentence. The combined system consistently finds suitable parsers and even outperforms the oracle single-best. The work shows how important a realistic experimental design is for truly low-resource languages.

Key points

  • Three distance measures:
    • KL-POS: the metric of Rosa and Žabokrtský (2015), but with linear interpolation smoothing following Brants (2000).
    • LANG-ID: langid.py, a naive Bayes classifier over byte n-grams (1–4), with pretrained models for ~100 languages.
    • WALS: Hamming distance between feature vectors; the database contains 2 679 languages and 202 features.
  • Example with a single Croatian sentence: KL-POS selects the Finnish parser (19th of 26), while with five sentences it already selects the Slovenian one (1st). LANG-ID and WALS select Slovenian already for a single sentence.
  • The test-set-level single-best oracle achieves a UAS of 51.96±4.35. If the best source were chosen separately for each sentence, UAS would rise to 65.54 (+13.58). In 28.66% of sentences, the best parse is produced by a parser other than the one ranked first overall.
  • Limitations: tokenization and sentence segmentation are taken from the gold annotation. The cross-lingual taggers of Agić et al. (2016) are not selected at run time but are used as data.
  • Code and data are openly available: the freasy repository on Bitbucket.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics: Part-of-speech tagging

Cited by

23 papers of our library cite it (of the 140673 whose bibliographies we read); 4 of them have a page here, 19 not yet.

Cite

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

@inproceedings{agic2017cross,
  title = {{Cross-Lingual Parser Selection for Low-Resource Languages}},
  author = {Agić, Željko},
  booktitle = {Proceedings of the NoDaLiDa 2017 Workshop on Universal Dependencies (UDW 2017)},
  year = {2017},
  url = {https://aclanthology.org/W17-0401/}
}

References

31 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Željko Agić, Dirk Hovy, and Anders Søgaard. 2015. If all you have is a bit of the Bible: Learning POS taggers for truly low-resource languages. In ACL, pages 268–272. find
  2. Plank and Van Noord (2011) applied similarity measures over cross-domain data for dependency Tahira Naseem, Regina Barzilay, and Amir Globerson. 2012. Selective Sharing for Multilingual Dependency Parsing. In ACL, pages 629–637. find
  3. Željko Agić, Anders Johannsen, Barbara Plank, Héctor Martı́nez Alonso, Natalie Schluter, and Anders Søgaard. 2016. Multilingual Projection for Parsing Truly Low-Resource Languages. TACL, 4:301–312. find
  4. Joakim Nivre, Marie-Catherine de Marneffe, Filip Ginter, Yoav Goldberg, Jan Hajič, Christopher D Manning, Ryan McDonald, Slav Petrov, Sampo Pyysalo, Natalia Silveira, et al. 2016. Universal Dependencies v1: A Multilingual Treebank Collection. In LREC 2016, pages 1659–1666. find our card
  5. Waleed Ammar, George Mulcaire, Miguel Ballesteros, Chris Dyer, and Noah A. Smith. 2016. Many Languages, One Parser. TACL, pages 431–444. find our card
  6. Bernd Bohnet. 2010. Very High Accuracy and Fast Dependency Parsing is not a Contradiction. In COL-ING, pages 89–97. find our card
  7. Thorsten Brants. 2000. TnT: A Statistical Part-of-Speech Tagger. In ANLP, pages 224–231. find our card
  8. Helen O’Horan, Yevgeni Berzak, Ivan Vulić, Roi Reichart, and Anna Korhonen. 2016. Survey on the Use of Typological Information in Natural Language Processing. arXiv preprint arXiv:1610.03349. find
  9. Dipanjan Das and Slav Petrov. 2011. Unsupervised Part-of-Speech Tagging with Bilingual Graph-Based Projections. In ACL, pages 600–609. find
  10. Robert Östling. 2015. Word Order Typology Through Multilingual Word Alignment. In ACL, pages 205– 211. find
  11. Slav Petrov, Dipanjan Das, and Ryan McDonald. 2011. A Universal Part-of-Speech Tagset. arXiv preprint arXiv:1104.2086. find
  12. Matthew S. Dryer and Martin Haspelmath, editors. 2013. The World Atlas of Language Structures Online. Max Planck Institute for Evolutionary Anthropology, Leipzig. find
  13. Barbara Plank and Gertjan Van Noord. 2011. Effective Measures of Domain Similarity for Parsing. In ACL, pages 1566–1576. find
  14. Ryan Georgi, Fei Xia, and William Lewis. 2010. Comparing Language Similarity Across Genetic and Typologically-Based Groupings. In COLING, pages 385–393. find
  15. Taraka Rama and Prasanth Kolachina. 2012. How Good are Typological Distances for Determining Genealogical Relationships among Languages? In COLING, pages 975–984. find
  16. Jan Hajič, Massimiliano Ciaramita, Richard Johansson, Daisuke Kawahara, Maria Antònia Martı́, Lluı́s Màrquez, Adam Meyers, Joakim Nivre, Sebastian Padó, Jan Štěpánek, et al. 2009. The CoNLL-2009 Shared Task: Syntactic and Semantic Dependencies in Multiple Languages. In CoNLL, pages 1–18. find our card
  17. Mohammad Sadegh Rasooli and Michael Collins. 2015. Density-Driven Cross-Lingual Transfer of Dependency Parsers. In EMNLP, pages 328–338. find
  18. Rudolf Rosa and Zdeněk Žabokrtský. 2015a. KLcpos3 - a Language Similarity Measure for Delexicalized Parser Transfer. In ACL, pages 243–249. find
  19. Anders Johannsen, Željko Agić, and Anders Søgaard. 2016. Joint Part-of-Speech and Dependency Projection from Multiple Sources. In ACL, pages 561–566. find
  20. Rudolf Rosa and Zdeněk Žabokrtský. 2015b. MST-Parser Model Interpolation for Multi-source Delexicalized Transfer. In IWPT, pages 71–75. find
  21. Marco Lui and Timothy Baldwin. 2012. langid.py: An Off-the-Shelf Language Identification Tool. In ACL, pages 25–30. find our card
  22. Kenji Sagae and Alon Lavie. 2006. Parser Combination by Reparsing. In NAACL, pages 129–132. find
  23. Xuezhe Ma and Fei Xia. 2014. Unsupervised Dependency Parsing with Transferring Distribution via Parallel Guidance and Entropy Regularization. In ACL, pages 1337–1348. find
  24. Anders Søgaard and Julie Wulff. 2012. An Empirical Study of Non-Lexical Extensions to Delexicalized Transfer. In COLING, pages 1181–1190. find
  25. Ryan McDonald, Slav Petrov, and Keith Hall. 2011. Multi-Source Transfer of Delexicalized Dependency Parsers. In EMNLP, pages 62–72. find our card
  26. Anders Søgaard. 2011. Data Point Selection for Cross-Language Adaptation of Dependency Parsers. In ACL, pages 682–686. find
  27. Ryan McDonald, Joakim Nivre, Yvonne Quirmbach-Brundage, Yoav Goldberg, Dipanjan Das, Kuzman Ganchev, Keith Hall, Slav Petrov, Hao Zhang, Oscar Täckström, Claudia Bedini, Núria Bertomeu Castelló, and Jungmee Lee. 2013. Universal Dependency Annotation for Multilingual Parsing. In ACL, pages 92–97. find
  28. Oscar Täckström, Ryan McDonald, and Joakim Nivre. 2013. Target Language Adaptation of Discriminative Transfer Parsers. In NAACL, pages 1061–1071. find
  29. Daniel Zeman and Philip Resnik. 2008. Cross-Language Parser Adaptation between Related Languages. In IJCNLP, pages 35–42. find
  30. Yuan Zhang and Regina Barzilay. 2015. Hierarchical Low-Rank Tensors for Multilingual Transfer Parsing. In EMNLP, pages 1857–1867. find
  31. Thomas Müller, Helmut Schmid, and Hinrich Schütze. 2013. Efficient Higher-Order CRFs for Morphological Tagging. In EMNLP, pages 322–332. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/agic-2017-cross-lingual-parser-selection-low/