Elamur Research⌕
☰
home / papers / 2010 / A Cascade Method for Detecting Hedges and their Scope in Natural Language Text

A Cascade Method for Detecting Hedges and their Scope in Natural Language Text

Summary

This paper describes a system built for the CoNLL-2010 shared task. The task has two parts. The first is deciding whether sentences contain hedges (uncertainty cues), using BioScope and Wikipedia weasel-word data. The second is finding each cue's in-sentence scope, using BioScope only. Hedged statements should not be treated as facts, so recognising them helps information extraction and inference.

The authors use a cascade:

  • Cue detection: cues are tagged with a B/I/O scheme. A CRF and a large-margin sequence labeller each make predictions. A second CRF then combines those outputs with lexical, POS and chunk features from GENIA Tagger.
  • Scope detection: a separate CRF labels the first and last token of each scope. It uses features from GDep Tagger, including dependency paths. Dictionary-based rules then repair incomplete scope boundaries.

The submitted system reached 86.36% F-measure for hedge detection on BioScope, which was the top biological result in the task. It scored 55.05% on Wikipedia and 49.95% for scope detection. In later revised experiments, a single CRF slightly outperformed the cascade in-domain.

Key points

  • Toolkits: CRF++ 0.53 and svm_hmm 3.10 (C=8000, tuned on a held-out third of the training data). Preprocessing used GENIA Tagger 3.0.1 for cues and GDep Tagger for scope.
  • Sentence-level hedge detection F1:
    • BioScope: cascade 86.36, single CRF 86.79, large margin 86.46.
    • Wikipedia: cascade 55.05, CRF 50.54. Recall on Wikipedia was low, around 36–41%.
  • Scope F1 with predicted cues was about 50, compared with 56.92 using gold cues. Scope-end tokens (L scope) were much harder than scope-start tokens (F scope): 74.14 vs 84.94 F1 with gold cues.
  • A cross-domain model trained on BioScope+Wikipedia performed worse than the in-domain models.
  • The scope representation follows Morante and Daelemans (2009). The reported numbers differ from the official submission because feature-extraction bugs were fixed afterwards.

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

18 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 16 not yet.

Cite

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

@inproceedings{tang2010cascade,
  title = {{A Cascade Method for Detecting Hedges and their Scope in Natural Language Text}},
  author = {Tang, Buzhou and Wang, Xiaolong and Wang, Xuan and Yuan, Bo and Fan, Shixi},
  booktitle = {Proceedings of the Fourteenth Conference on Computational Natural Language Learning – Shared Task},
  year = {2010},
  url = {https://aclanthology.org/W10-3002/}
}

References

5 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Richárd Farkas, Veronika Vincze, György Móra, János Csirik, György Szarvas. 2010. The CoNLL-2010 Shared Task: Learning to Detect Hedges and their Scope in Natural Language Text. In Proceedings of the Fourteenth Conference on Computational Natural Language Learning (CoNLL-2010): Shared Task, pages 1–12, Uppsala, Sweden, July. Association for Computational Linguistics. find our card
  2. Viola Ganter and Michael Strube. 2009. Finding hedges by chasing weasels: Hedge detection using wikipedia tags and shallow linguistic features. In Proceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 173–176, Suntec, Singapore, August. Association for Computational Linguistics. find our card
  3. Roser Morante and Walter Daelemans. 2009. Learning the scope of hedge cues in biomedical texts. In Proceedings of the BioNLP 2009 Workshop, pages 28– 36, Boulder, Colorado, June. Association for Computational Linguistics. find our card
  4. Tzong-Han Tsai, Chia-Wei Wu, and Wen-Lian Hsu. 2005. Using Maximum Entropy to Extract Biomedical Named Entities without Dictionaries. In Second International Joint Conference on Natural Language Processing, pages 268–273. find
  5. Veronika Vincze, György Szarvas, Richárd Farkas, György Móra, and János Csirik. 2008. The Bio-Scope corpus: biomedical texts annotated for uncertainty, negation and their scopes. BMC Bioinformatics, 9(Suppl 11):S9. find

This page in Markdown → https://md.research.elamur.ai/free/en/papers/tang-2010-cascade-method-detecting-hedges-their/