Elamur Research⌕
☰
home / papers / 2018 / Identifying Explicit Discourse Connectives in German

Identifying Explicit Discourse Connectives in German

Summary

The paper addresses the first stage of a pipeline-style shallow discourse parser for German. That stage decides whether a candidate word from the DiMLex lexicon acts as an explicit discourse connective or has only a sentential reading. Annotated German data is scarce, so the authors use the Potsdam Commentary Corpus (PCC), a collection of 176 newspaper editorials. They reuse the surface, part-of-speech and syntactic features of the English system by Lin et al. (2014), train a Random Forest, and evaluate with 10-fold cross-validation. Error analysis led them to add three features: a coarse syntactic category for the candidate, whether it opens an S/Sbar clause, and sentence length. They also tried to enlarge the training set by projecting English classifier output onto German through EuroParl word alignments, but this lowered performance. The extended model reaches an f-score of 83.89, roughly four points above a majority baseline. A learning-curve comparison suggests German is no harder than English when data volumes match. More in-domain annotation therefore looks like the most promising next step.

Key points

  • Data: PCC has 33,222 words and 1,176 connective instances, covering 156 unique connectives (DiMLex lists 275 entries). In total, 3,406 candidate instances were extracted.
  • F-scores (micro-averaged):
    • Majority baseline: 79.60
    • Base features with automatic parses: 81.90
    • Base features with gold PCC trees: 85.76
    • Extended feature set: 83.89
    • Extended set with prepositions excluded: 85.99
  • Annotation projection: 18,853 extra instances were obtained from EuroParl using MGIZA alignments. They pushed the f-score down to about 66 (65.96 in Table 1), close to the 68.7 reported by Versley (2010).
  • English comparison: With the same features and classifier, English trained on a comparably small sample scores 80.09. On the full CoNLL 2016 training data it scores 93.64.
  • Difficult cases:
    • 'auch' and 'als' were hardest, with f-scores of 27.03 and 28.57.
    • With separate classifiers per syntactic category, prepositions scored worst (51.55).
    • Coordinating and subordinating conjunctions scored best (91.81 and 90.25).

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: Discourse analysis

Cited by

5 papers of our library cite it (of the 140673 whose bibliographies we read); 3 of them have a page here, 2 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/w18-5037

@inproceedings{bourgonje2018identifying,
  title = {{Identifying Explicit Discourse Connectives in German}},
  author = {Bourgonje, Peter and Stede, Manfred},
  booktitle = {Proceedings of the 19th Annual SIGdial Meeting on Discourse and Dialogue},
  year = {2018},
  doi = {10.18653/v1/w18-5037},
  url = {https://doi.org/10.18653/v1/w18-5037}
}

References

16 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Or Biran and Kathleen McKeown. 2015. PDTB Discourse Parsing as a Tagging Task: The Two Taggers Approach. In SIGDIAL Conference. The Association for Computer Linguistics, pages 96–104. find our card
  2. Tatjana Scheffler and Manfred Stede. 2016. Adding semantic relations to a large-coverage connective lexicon of German. In Nicoletta Calzolari et al., editor, Proc. of the Ninth International Conference on Language Resources and Evaluation (LREC 2016). Portoro, Slovenia. find
  3. Stefanie Dipper and Manfred Stede. 2006. Disambiguating potential connectives. In Proceedings of the KONVENS Conference. Konstanz. find
  4. Qin Gao and Stephan Vogel. 2008. Parallel implementations of word alignment tool. In Software Engineering, Testing, and Quality Assurance for Natural Language Processing. Association for Computational Linguistics, SETQA-NLP ’08, pages 49–57. find
  5. Radu Soricut and Daniel Marcu. 2003. Sentence level discourse parsing using syntactic and lexical information. In Proceedings of the 2003 Human Language Technology Conference of the North American Chapter of the Association for Computational Linguistics. find our card
  6. Philipp Koehn. 2005. Europarl: A parallel corpus for statistical machine translation. In Conference Proceedings: the tenth Machine Translation Summit. AAMT, Phuket, Thailand, pages 79–86. find
  7. Manfred Stede. 2002. DiMLex: A lexical approach to discourse markers. In Exploring the Lexicon - Theory and Computation, Edizioni dell’Orso, Alessandria. find
  8. Zhongyi Li, Hai Zhao, Chenxi Pang, Lili Wang, and Huan Wang. 2016. A Constituent Syntactic Parse Tree Based Discourse Parser. In Proceedings of the CONLL 2016 Shared Task. Berlin, pages 60–64. find our card
  9. Manfred Stede and Arne Neumann. 2014. Potsdam Commentary Corpus 2.0: Annotation for discourse research. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14). European Language Resources Association (ELRA), Reykjavik, Iceland. find
  10. Ziheng Lin, Hwee Tou Ng, and Min-Yen Kan. 2014. A PDTB-Styled End-to-End Discourse Parser. Natural Language Engineering 20:151–184. find
  11. Stephan Oepen, Jonathon Read, Tatjana Scheffler, Uladzimir Sidarenka, Manfred Stede, Erik Velldal, and Lilja Øvrelid. 2016. OPT: OsloPotsdamTeesside—Pipelining Rules, Rankers, and Classifier Ensembles for Shallow Discourse Parsing. In Proceedings of the CONLL 2016 Shared Task. Berlin. find
  12. Yannick Versley. 2010. Discovery of Ambiguous and Unambiguous Discourse Connectives via Annotation Projection. In Proceedings of Workshop on Annotation and Exploitation of Parallel Corpora (AEPC). Northern European Association for Language Technology (NEALT). ↗
  13. Renate Pasch, Ursula Brauße, Eva Breindl, and Ulrich Herrmann Waßner. 2003. Handbuch der deutschen Konnektoren. Walter de Gruyter, Berlin/New York. ↗
  14. Yannick Versley and Anna Gastel. 2012. Linguistic tests for discourse relations in the TüBa-D/Z corpus of written German. Dialogue and Discourse pages 1–24. find our card
  15. Fabian Pedregosa, Gael 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 Edouard Duchesnay. 2011. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research 12:2825–2830. ↗
  16. Jianxiang Wang and Man Lan. 2015. A Refined End-to-End Discourse Parser. In Proceedings of the Nineteenth Conference on Computational Natural Language Learning - Shared Task. Association for Computational Linguistics, pages 17–24. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/bourgonje-2018-identifying-explicit-discourse-connectives-german/