Elamur Research⌕
☰
home / papers / 2020 / Shallow Discourse Parsing for Under-Resourced Languages: Combining Machine Translation and Annotation Projection

Shallow Discourse Parsing for Under-Resourced Languages: Combining Machine Translation and Annotation Projection

Summary

Training shallow discourse parsers requires large annotated corpora, and only English has one at sufficient scale: the Penn Discourse TreeBank (PDTB). The authors therefore build a German resource automatically. They machine-translate the English PDTB 2.0 (the CoNLL 2016 shared-task subset) into German and transfer its annotations onto the translation via word alignment. Unlike earlier projection work, which handled only explicit connectives, this approach transfers whole coherence relations, including arguments, senses and non-explicit types. Alignment comes from GIZA++, refined with Moses symmetrisation heuristics, and connective candidates are checked against the DiMLex lexicon. The result is GermanPDTB, a silver-standard corpus of German financial news. It is far larger than the existing gold German data and also yields new candidate entries for DiMLex. The authors analyse translation and projection errors, then test the corpus extrinsically: they train selected parser components on GermanPDTB and compare them with the same components trained on the gold English PDTB. The comparison figures are not included in the excerpt provided. The work matters because it offers a reusable recipe for any language with decent MT from English.

Key points

  • Source corpus vs. German data: PDTB 2.0 has over 1M words and more than 40k relations. The largest gold German corpus, the Potsdam Commentary Corpus (PCC), has only 2,208 relations.
  • Choice of MT system: Five systems were compared by BLEU on an OPUS news corpus (Google Translate, DeepL, Bing, Edinburgh NMT, Moses). DeepL (28.07) and Google (26.6) scored best. A 50-sentence manual check (Popovic et al., 2013) rated them as equal in quality. DeepL was chosen because its alignments recovered more explicit connectives.
  • Alignment method: Six alignment variants were tested (intersection, union, grow, grow-diag, grow-diag-final, grow-diag-final-and) by matching aligned connectives against DiMLex. Intersection became the default because the authors prioritised precision over recall.
  • Relations projected: Explicit (~43%), Implicit (~38%), EntRel (~12%) and AltLex (~2%) relations were projected. NoRel (~0.6%) was excluded.

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 · Machine translation

Topics: Discourse analysis · Machine translation

Cited by

10 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 8 not yet.

Cite

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

@inproceedings{sluytergathje2020shallow,
  title = {{Shallow Discourse Parsing for Under-Resourced Languages: Combining Machine Translation and Annotation Projection}},
  author = {Sluyter-Gäthje, Henny and Bourgonje, Peter and Stede, Manfred},
  booktitle = {Proceedings of the Twelfth Language Resources and Evaluation Conference},
  year = {2020},
  url = {https://aclanthology.org/2020.lrec-1.131/}
}

References

29 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Asher, N. and Lascarides, A. (2005). Logics of Conversation. Studies in Natural Language Processing. Cambridge University Press. find
  2. Bourgonje, P. and Stede, M. (2018). Identifying Explicit Discourse Connectives in German. In Proceedings of the 19th Annual SIGdial Meeting on Discourse and Dialogue, pages 327–331, Melbourne, Australia. Association for Computational Linguistics. find our card
  3. Bourgonje, P. and Stede, M. (2019). Explicit Discourse Argument Extraction for German. In Proceedings of the 21st International Conference on Text, Speech and Dialogue, Ljubljana, Slovenia. find
  4. Bourgonje, P. and Stede, M. (2020). The Potsdam Commentary Corpus 2.2: Extending Annotations for Shallow Discourse Parsing. In Proceedings of the 12th International Conference on Language Resources and Evaluation (LREC’20), Paris, France, May. European Language Resources Association (ELRA). find our card
  5. Bourgonje, P., Grishina, Y., and Stede, M. (2017). Toward a bilingual lexical database on connectives: Exploiting a German/Italian parallel corpus. In Proceedings of the Fourth Italian Conference on Computational Linguistics, Rome, Italy, December. find
  6. Hajlaoui, N. and Popescu-Belis, A. (2013). Assessing the Accuracy of Discourse Connective Translations: Validation of an Automatic Metric. In 14th International Conference on Intelligent Text Processing and Computational Linguistics. Springer. find
  7. Koehn, P., Hoang, H., Birch, A., Callison-Burch, C., Federico, M., Bertoldi, N., Cowan, B., Shen, W., Moran, C., Zens, R., Dyer, C., Bojar, O., Constantin, A., and Herbst, E. (2007). Moses: Open Source Toolkit for Statistical Machine Translation. In Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics Companion, pages 177–180, Prague, Czech Republic, June. Association for Computational Linguistics. find
  8. Laali, M. and Kosseim, L. (2014). Inducing Discourse Connectives from Parallel Texts. In Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 610–619, Dublin, Ireland, August. Dublin City University and Association for Computational Linguistics. find our card
  9. Laali, M. (2017). Inducing Discourse Resources Using Annotation Projection. Ph.D. thesis, Concordia University. find
  10. Li, Z., Zhao, H., Pang, C., Wang, L., and Wang, H. (2016). A Constituent Syntactic Parse Tree Based Discourse Parser. In Proceedings of the CoNLL-16 shared task, pages 60–64, Berlin, Germany, August. Association for Computational Linguistics. find our card
  11. Mann, W. and Thompson, S. (1988). Rhetorical Structure Theory: Towards a Functional Theory of Text Organization. Text, 8:243–281. find
  12. Meyer, T. and Webber, B. (2013). Implicitation of Discourse Connectives in (Machine) Translation. In Proceedings of the Workshop on Discourse in Machine Translation, pages 19–26. find our card
  13. Müller, T., Schmid, H., and Schütze, H. (2013). Efficient higher-order CRFs for morphological tagging. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 322–332. find
  14. Och, F. J. and Ney, H. (2003). A Systematic Comparison of Various Statistical Alignment Models. Computational Linguistics, 29(1):19–51. ↗
  15. Oepen, S., Read, J., Scheffler, T., Sidarenka, U., Stede, M., Velldal, E., and Øvrelid, L. (2016). OPT: Oslo–Potsdam–Teesside—Pipelining Rules, Rankers, and Classifier Ensembles for Shallow Discourse Parsing. In Proceedings of the CONLL 2016 Shared Task, Berlin. find our card
  16. Papineni, K., Roukos, S., Ward, T., and Zhu, W.-J. (2002). Bleu: a Method for Automatic Evaluation of Machine Translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, pages 311–318, Philadelphia, Pennsylvania, USA, July. Association for Computational Linguistics. find our card
  17. Popovic, M., Avramidis, E., Burchardt, A., Hunsicker, S., Schmeier, S., Tscherwinka, C., Vilar, D., and Uszkoreit, H. (2013). Learning from human judgments of machine translation output. In Proceedings of the MT Summit XIV. Machine Translation Summit (MT-Summit-2013), Nice, France, September. The European Association for Machine Translation. find
  18. Prasad, R., Dinesh, N., Lee, A., Joshi, A., and Webber, B. (2006). Annotating Attribution in the Penn Discourse TreeBank. In Proceedings of the Workshop on Sentiment and Subjectivity in Text, SST ’06, pages 31–38, Stroudsburg, PA, USA. Association for Computational Linguistics. find our card
  19. Prasad, R., Dinesh, N., Lee, A., Miltsakaki, E., Robaldo, L., Joshi, A., and Webber, B. (2008). The Penn Discourse TreeBank 2.0. In Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC’08), Marrakech, Morocco, May. European Language Resources Association (ELRA). find our card
  20. Sennrich, R., Haddow, B., and Birch, A. (2016). Edinburgh Neural Machine Translation Systems for WMT 16. In Proceedings of the First Conference on Machine Translation: Volume 2, Shared Task Papers, pages 371– 376, Berlin, Germany, August. Association for Computational Linguistics. find our card
  21. Stede, M. (2002). DiMLex: A Lexical Approach to Discourse Markers. In Lenci A. et al., editors, Exploring the Lexicon - Theory and Computation. Edizioni dell’Orso, Alessandria. find
  22. Tiedemann, J. (2012). Parallel data, tools and interfaces in OPUS. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC’12), pages 2214–2218, Istanbul, Turkey, May. European Language Resources Association (ELRA). find
  23. Versley, Y. and Gastel, A. (2012). Linguistic Tests for Discourse Relations in the TüBa-D/Z corpus of written German. Dialogue and Discourse, 4(2):142–173. find our card
  24. Versley, Y. (2010). Discovery of ambiguous and unambiguous discourse connectives via annotation projection. In L. Ahrenberg, et al., editors, Proceedings of Workshop on Annotation and Exploitation of Parallel Corpora (AEPC), pages 83–92. Northern European Association for Language Technology (NEALT). ↗
  25. Xue, N., Ng, H. T., Pradhan, S., Rutherford, A., Webber, B., Wang, C., and Wang, H. (2016). CoNLL 2016 Shared Task on Multilingual Shallow Discourse Parsing. In Proceedings of the CoNLL-16 shared task, pages 1– 19, Berlin, Germany, August. Association for Computational Linguistics. find our card
  26. Zeldes, A., Das, D., Maziero, E. G., Antonio, J., and Iruskieta, M. (2019). find
  27. The DISRPT 2019 Shared Task on Elementary Discourse Unit Segmentation and Connective Detection. In Proceedings of the Workshop on Discourse Relation Parsing and Treebanking 2019, pages 97–104, Minneapolis, MN, June. Association for Computational Linguistics. find our card
  28. Zhou, L., Gao, W., Li, B., Wei, Z., and Wong, K.-F. (2012). Cross-Lingual Identification of Ambiguous Discourse Connectives for Resource-Poor Language. In Proceedings of COLING 2012: Posters, pages 1409–1418, Mumbai, India, December. find our card
  29. The COLING 2012 Organizing Committee. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/gathje-2020-shallow-discourse-parsing-under-resourced/