Elamur Research⌕
☰
home / papers / 2016 / IMS HotCoref DE: A Data-driven Co-reference Resolver for German

IMS HotCoref DE: A Data-driven Co-reference Resolver for German

Summary

German coreference resolution has received much less attention than English. After the SemEval 2010 shared task, only a few German systems appeared, and these were mostly rule-based or tied to a specific domain. The authors address this gap by porting the English IMS HotCoref resolver to German. That resolver treats a document's coreference structure as a rooted tree and learns with latent antecedents and non-local features. The learning algorithm stays the same. The adaptation work focuses on preprocessing and on features for German-specific difficulties: rich inflection, grammatical gender on inanimate nouns, and single-word compounds. Lexical-semantic knowledge comes from GermaNet and from word embeddings. The system is trained and tested on the TüBa-D/Z newspaper treebank. With gold annotations it reaches a CoNLL score of 65.76 on version 10. In a post-task SemEval 2010 evaluation it outperforms the previous best system, CorZu. Ablation experiments indicate that the linguistically motivated features help. The tool is released for free download, which makes it a practical, reproducible baseline for German coreference.

Key points

  • Data and splits: The system uses TüBa-D/Z, a gold-annotated newspaper corpus of about 1.8M tokens.
    • Version 8 is used with the official SemEval splits.
    • Version 10 is split by the authors in a 20-20-60 test-dev-train ratio (727/727/2190 documents).
  • Results on version 10: The CoNLL score is 65.76 with gold annotations and 48.54 with predicted ones. Mention extraction recall is about 99% with gold input versus about 78% with predicted input, with most misses caused by parsing errors.
  • SemEval 2010 post-task (TüBa 8): The system scores 63.61 in the open gold track and 48.61 in the open regular track. CorZu scores 58.11 and 45.82, and the gap is statistically significant.
  • German-specific features:
    • All features are based on lemmas rather than word forms.
    • A gender agreement feature is added.
    • Compound head matching is implemented in two ways: a simple suffix heuristic and a version based on the COMPOST splitter.
    • GermaNet is used to look up synonyms, hypernyms and hyponyms.
    • Person and name lists from Klenner and Tuggener (2011) supply animacy and name-gender information.
  • Distributional feature: A CBOW model is trained on lemmatised SdeWaC (tagged with TreeTagger) using lemma-POS pairs, 300 dimensions and negative sampling of 15. Two mentions count as similar when their cosine similarity, computed with DISSECT, is at least 0.8.

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: Coreference resolution

Cited by

9 papers of our library cite it (of the 140673 whose bibliographies we read); 3 of them have a page here, 6 not yet.

Cite

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

@inproceedings{roesiger2016ims,
  title = {{IMS HotCoref DE: A Data-driven Co-reference Resolver for German}},
  author = {Roesiger, Ina and Kuhn, Jonas},
  booktitle = {Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16)},
  year = {2016},
  doi = {10.63317/5d86anb3gozo},
  url = {https://doi.org/10.63317/5d86anb3gozo}
}

References

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

  1. Attardi, G., Simi, M., and Dei Rossi, S. (2010). Tanl-1: Coreference resolution by parse analysis and similarity clustering. In Proceedings of the 5th International Workshop on Semantic Evaluation, pages 108–111, Uppsala, Sweden, July. Association for Computational Linguistics. find our card
  2. Bergsma, S. and Lin, D. (2006). Bootstrapping path-based pronoun resolution. In Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, pages 33–40, Sydney, Australia, July. Association for Computational Linguistics. find our card
  3. Björkelund, A. and Kuhn, J. (2014). Learning structured perceptrons for coreference resolution with latent antecedents and non-local features. In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 47– 57, Baltimore. find our card
  4. Bohnet, B. and Nivre, J. (2012). A transition-based system for joint part-of-speech tagging and labeled non-projective dependency parsing. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 1455–1465, Jeju Island, Korea, July. Association for Computational Linguistics. find
  5. Broscheit, S., Poesio, M., Ponzetto, S. P., Rodriguez, K. J., Romano, L., Uryupina, O., Versley, Y., and Zanoli, R. (2010a). Bart: A multilingual anaphora resolution system. In Proceedings of the 5th International Workshop on Semantic Evaluation, pages 104–107, Uppsala, Sweden, July. Association for Computational Linguistics. find our card
  6. Broscheit, S., Ponzetto, S. P., Versley, Y., and Poesio, M. (2010b). Extending BART to provide a coreference resolution system for german. In Proceedings of the International Conference on Language Resources and Evaluation, LREC 2010, 17-23 May 2010, Valletta, Malta. find our card
  7. Cap, F. (2014). Morphological processing of compounds for statistical machine translation. Dissertation, Institute for Natural Language Processing (IMS), University of Stuttgart. ↗
  8. Clark, K. and Manning, C. D. (2015). Entity-centric coreference resolution with model stacking. In Association of Computational Linguistics (ACL). find our card
  9. Dinu, G., Pham, N. T., and Baroni, M. (2013). DISSECT – DIStributional SEmantics Composition Toolkit. In Proceedings of ACL, Sofia, Bulgaria. find
  10. Durrett, G. and Klein, D. (2014). A joint model for entity analysis: Coreference, typing, and linking. Transactions of the Association for Computational Linguistics, 2:477– 490. ↗ our card
  11. Faaß, G. and Eckart, K. (2013). SdeWaC – a corpus of parsable sentences from the web. In Language Processing and Knowledge in the Web, Lecture Notes in Computer Science. Springer. find
  12. Faruqui, M. and Padó, S. (2010). Training and evaluating a german named entity recognizer with semantic generalization. In Proceedings of KONVENS 2010, Saarbrücken, Germany. find
  13. Klenner, M. and Tuggener, D. (2011). An incremental entity-mention model for coreference resolution with restrictive antecedent accessibility. In Proceedings of RANLP, pages 178–185, Hissar, Bulgaria. find our card
  14. Kobdani, H. and Schütze, H. (2010). Sucre: A modular system for coreference resolution. In Proceedings of the 5th International Workshop on Semantic Evaluation, pages 92–95, Uppsala, Sweden, July. Association for Computational Linguistics. find our card
  15. Krug, M., Puppe, F., Jannidis, F., Macharowsky, L., Reger, I., and Weimer, L. (2015). Rule-based coreference resolution in german historic novels. In Computational Linguistics for Literature. find our card
  16. Mikhaylova, A. (2014). Koreferenzresolution in mehreren sprachen. Msc thesis, Center for Information and Language Processing, University of Munich. find
  17. Mikolov, T., Chen, K., Corrado, G., and Dean, J. (2013). Efficient estimation of word representations in vector space. In Proceedings of ICLR, Scottsdale, AZ, USA. ↗
  18. Naumann, K. (2006). Manual for the annotation of in-document referential relations. University of Tübingen. find
  19. Ng, V. (2010). Supervised noun phrase coreference research: The first fifteen years. In Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages 1396–1411. find our card
  20. Petrov, S., Barrett, L., Thibaux, R., and Klein, D. (2006). Learning accurate, compact, and interpretable tree annotation. In Proceedings of the 21st International Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics, pages 433–440. Association for Computational Linguistics. find our card
  21. Pradhan, S., Ramshaw, L., Marcus, M., Palmer, M., Weischedel, R., and Xue, N. (2011). CoNLL-2011 Shared Task: Modeling unrestricted coreference in OntoNotes. In Proceedings of the Fifteenth Conference on Computational Natural Language Learning: Shared Task, pages 1–27, Stroudsburg, PA, USA. find our card
  22. Pradhan, S., Moschitti, A., Xue, N., Uryupina, O., and Zhang, Y. (2012). Conll-2012 shared task: Modeling multilingual unrestricted coreference in ontonotes. In Joint Conference on EMNLP and CoNLL-Shared Task, pages 1–40. Association for Computational Linguistics. find our card
  23. Recasens, M., Màrquez, L., Sapena, E., Martı́, M. A., Taulé, M., Hoste, V., Poesio, M., and Versley, Y. (2010). Semeval-2010 task 1: Coreference resolution in multiple languages. In Proceedings of the 5th International Workshop on Semantic Evaluation, SemEval ’10, pages 1–8, Stroudsburg, PA, USA. Association for Computational Linguistics. find our card
  24. Schmid, H. (1994). Probabilistic part-of-speech tagging using decision trees. In Proceedings of NeMLaP, Manchester, UK. find
  25. Tuggener, D. and Klenner, M. (2014). A hybrid entity-mention pronoun resolution model for german using markov logic networks. In Proceedings of KONVENS 2014, pages 21–29. ↗
  26. Zhekova, D. and Kübler, S. (2010). Ubiu: A language-independent system for coreference resolution. In Proceedings of the 5th International Workshop on Semantic Evaluation, pages 96–99, Uppsala, Sweden, July. Association for Computational Linguistics. find our card
  27. Ziering, P. (2011). Feature engineering for coreference resolution in german: Improving the link feature set of sucre for german by using a more linguistic background. Diploma thesis, Institute for Natural Language Processing, University of Stuttgart. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/roesiger-2016-ims-hotcoref-de-data-driven/