---
title: "Exploiting Rich Features for Detecting Hedges and their Scope"
description: "This CoNLL-2010 shared task entry tackles two linked problems in biomedical text. 0.5265 F1. 12 references, cited by 4 papers here."
authors: ["Xinxin Li", "Jianping Shen", "Xiang Gao", "Xuan Wang"]
year: 2010
venue: "Proceedings of the Fourteenth Conference on Computational Natural Language Learning – Shared Task"
source: "https://aclanthology.org/W10-3011/"
licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)"
licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/li-2010-exploiting-rich-features-detecting-hedges/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/li-2010-exploiting-rich-features-detecting-hedges/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2010](https://md.research.elamur.ai/free/en/papers/2010/page/2/) / Exploiting Rich Features for Detecting Hedges and their Scope

# Exploiting Rich Features for Detecting Hedges and their Scope

**Authors:** Xinxin Li, Jianping Shen, [Xiang Gao](https://md.research.elamur.ai/free/en/authors/xiang-gao/) and [Xuan Wang](https://md.research.elamur.ai/free/en/authors/xuan-wang/) · **Year:** 2010 · **Venue:** [Proceedings of the Fourteenth Conference on Computational Natural Language Learning – Shared Task](https://md.research.elamur.ai/free/en/venues/conll/) ([CoNLL 2010](https://md.research.elamur.ai/free/en/venues/conll/2010/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/)

**Links:** [Publisher's page](https://aclanthology.org/W10-3011/) · [PDF](https://aclanthology.org/W10-3011.pdf) · [OpenAlex](https://openalex.org/W1675612499)

**Citations:** 17 citations (OpenAlex, 2026-10-14) · [4](https://md.research.elamur.ai/free/en/papers/li-2010-exploiting-rich-features-detecting-hedges/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
This CoNLL-2010 shared task entry tackles two linked problems in biomedical text. The first is finding hedge cues, words or phrases that signal speculation. The second is finding the span of text each cue governs. The authors treat both as sequence labeling with CRFs, building on Morante and Daelemans (2009).

- **Cue detection:** tokens get BIO tags. Features (word form, lemma, POS, chunk tags from the GENIA tagger) are chosen by a greedy add/remove search scored on the development set. Hand-written chunk-based rules then repair missed or partial cues.
- **Scope finding:** a second CRF marks only the first and last tokens of each scope. Its features include tree paths to the nearest cue, taken from a GENIA dependency parser and a Brown self-trained biomedical constituency parser. A rule-based step then pairs boundaries with cues.

The system placed third on the biological cue task and reached 0.5265 F1 on scope finding. Syntax clearly helped on abstracts but not on full articles. This suggests structural features do not transfer equally across genres.

## Key points
- Both tasks use CRF++ 0.51 with default parameters. In cue detection, CRF beat SVM-HMM on both precision and recall.
- Cue post-processing raised I-cue recall from 55.26% to 68.51%. It cost a little precision and gave a net F1 gain of 0.44%.
- On abstracts, phrase-structure features added about 1.2% F1 over sequence-only features. All syntactic features together added 2.19%.
- On full articles with gold cues, scope F1 depended on the training data:
  - full-article data only: 52.46
  - abstracts only: 54.48
  - both combined: 59.66
- Dependency features slightly hurt scope finding on full articles. Without them, F1 was 59.52 with gold cues and 52.65 with predicted cues.

## Interesting ideas
- Instead of tagging every token inside a scope, the model predicts only the scope's start and end, and a deterministic step enforces global constraints (one scope per cue, cue inside its scope, nesting allowed but no crossing).
- The relation between a candidate token and a trigger word is described by its tree path: the POS chain (also with repeated tags collapsed), the common ancestor's POS, up/down direction, kinship type, and path length.
- Features are selected greedily by trying each addition and each removal against development-set performance until nothing improves.
- Multiword cues are completed with chunk-level consistency rules: if part of a noun phrase is tagged as a cue, the whole phrase is pulled in.
- Syntactic features that helped on abstracts were useless on full articles, which shows that feature gains must be checked per genre.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (frames): Modality and negation scope over frames could be resolved the same way, with boundaries predicted locally and well-formedness enforced by explainable rules.
- **I2** (tree): These path descriptors could serve as transparent features for attaching hedges, negators and quantifiers to the UD subtrees they govern.
- **I3** (learn): A cheap, auditable procedure like this could prune feature templates in the perceptron parser while recording why each template survived.
- **I4** (tokens): A similar span-closure rule could make multiword expressions like "raises a question" or "it is likely that" atomic units before parsing.
- **I5** (learn): Evaluations should report results separately for fables, school texts and word problems, so a gain in one register is not assumed to carry over.

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

## Institutions

- [Harbin Institute of Technology](https://md.research.elamur.ai/free/en/universities/harbin-institute-of-technology/) — University, Harbin, China · 76 works · 27,918 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works

## Cite

Cite as: Xinxin Li, Jianping Shen, Xiang Gao and Xuan Wang. 2010. Exploiting Rich Features for Detecting Hedges and their Scope. Proceedings of the Fourteenth Conference on Computational Natural Language Learning – Shared Task.

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

```bibtex
@inproceedings{li2010exploiting,
  title = {{Exploiting Rich Features for Detecting Hedges and their Scope}},
  author = {Xinxin Li, Jianping Shen, Xiang Gao and Xuan Wang},
  booktitle = {Proceedings of the Fourteenth Conference on Computational Natural Language Learning – Shared Task},
  year = {2010},
  url = {https://aclanthology.org/W10-3011/}
}
```

- [← previous Exploring the Data-Driven Prediction of Prepositions in English](https://md.research.elamur.ai/free/en/papers/elghafari-2010-exploring-data-driven-prediction-prepositions/)
- [next → Modeling Morphosyntactic Agreement in Constituency-Based Parsing of Modern Hebrew](https://md.research.elamur.ai/free/en/papers/tsarfaty-2010-modeling-morphosyntactic-agreement-constituency-based/)
- [↑ back to papers of 2010](https://md.research.elamur.ai/free/en/papers/2010/page/2/)

## References

12 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. David B. Aronow, Fangfang Feng, and W. Bruce Croft. 1999. Ad Hoc Classification of Radiology Reports. Journal of the American Medical Informatics Association , 6(5):393–411. [↗](https://doi.org/10.1136/jamia.1999.0060393)
2. Wendy W. Chapman, Will Bridewell, Paul Hanbury, Gregory F. Cooper, and Bruce G. Buchanan. 2001. A Simple Algorithm for Identifying Negated Findings and Diseases in Discharge Summaries. Journal of Biomedical Informatics , 34:301–310. [↗](https://doi.org/10.1006/jbin.2001.1029)
3. Richárd Farkas, Veronika Vincze, György Móra, János Csirik, and György Szarvas. 2010. [The CoNLL-2010 Shared Task: Learning to Detect Hedges and their Scope in Natural Language Text](https://md.research.elamur.ai/free/en/papers/farkas-2010-conll-2010-shared-task-learning/). In Proceedings of the Fourteenth Conference on Computational Natural Language Learning (CoNLL-2010): Shared Task , pages 1–12.
4. Yang Huang, and Henry J. Lowe. 2007. A novel hybrid approach to automated negation detection in clinical radiology reports. Journal of the American Medical Informatics Association , 14(3):304–311. F1 52. [↗](https://doi.org/10.1197/jamia.m2284)
5. John Lafferty, Andrew McCallum, and Fernando Pereira. 2001. Conditional random fields: 58. Probabilistic models for segmenting and labeling sequence data. In Proceedings of the Eighteenth International Conference Learning , pages 282–289. on Machine David McClosky. 2009. Any Domain Parsing: Automatic Domain Adaptation for Natural Language Parsing . Ph.D. thesis, Department of Computer Science, Brown University. [find](https://scholar.google.com/scholar?q=John+Lafferty%2C+Andrew+McCallum%2C+and+Fernando+Pereira%2E+2001%2E+Conditional+random+fields%3A+58%2E+Probabilistic+models+for+segmenting+and+labeling+sequence+data%2E+In+Proceedings+of+the+Eighteenth+Internationa)
6. Ben Medlock, and Ted Briscoe. 2007. [Weakly supervised learning for hedge classification in scientific literature](https://md.research.elamur.ai/free/en/papers/medlock-2007-weakly-supervised-learning-hedge-classification/). In Proc. of ACL 2007 , pages 992–999.
7. Roser Morante, and Walter Daelemans. 2009. [Learning the scope of hedge cues in biomedical texts](https://md.research.elamur.ai/free/en/papers/morante-2009-learning-scope-hedge-cues-biomedical/). In Proceedings of the Workshop on BioNLP , pages 28–36.
8. Pradeep G. Mutalik, Aniruddha Deshpande, and Prakash M. Nadkarni. 2001. Use of general-purpose negation detection to augment concept indexing of medical documents: a quantitative study using the UMLS. Journal of the American Medical Informatics Association , 8(6):598–609. [↗](https://doi.org/10.1136/jamia.2001.0080598)
9. Kenji Sagae, and Jun’ichi Tsujii. 2007. [Dependency parsing and domain adaptation with LR models and parser ensembles](https://md.research.elamur.ai/free/en/papers/sagae-2007-dependency-parsing-domain-adaptation-lr/). In Proceedings of the CoNLL-2007 Shared Task , pages 82–94
10. György Szarvas. 2008. [Hedge classification in biomedical texts with a weakly supervised selection of keywords](https://md.research.elamur.ai/free/en/papers/szarvas-2008-hedge-classification-biomedical-texts-weakly/). In Proc. of ACL 2008 , pages 281–289, Columbus, Ohio, USA. ACL.
11. György Szarvas, Veronika Vincze, Richárd Farkas, and János Csirik. 2008. [The BioScope corpus: annotation for negation, uncertainty and their scope in biomedical texts](https://md.research.elamur.ai/free/en/papers/szarvas-2008-bioscope-corpus-annotation-negation-uncertainty/). In Proc. of BioNLP 2008 , pages 38–45, Columbus, Ohio. ACL.
12. Yoshimasa Tsuruoka, Yuka Tateishi, Jin-Dong Kim, Tomoko Ohta, John McNaught, Sophia Ananiadou, and Jun'ichi Tsujii. 2005. Developing a Robust Part-of-Speech Tagger for Biomedical Text. Advances in Informatics - 10th Panhellenic Conference on Informatics, LNCS 3746, pages 382–392. [find](https://scholar.google.com/scholar?q=Yoshimasa+Tsuruoka%2C+Yuka+Tateishi%2C+Jin%2DDong+Kim%2C+Tomoko+Ohta%2C+John+McNaught%2C+Sophia+Ananiadou%2C+and+Jun%27ichi+Tsujii%2E+2005%2E+Developing+a+Robust+Part%2Dof%2DSpeech+Tagger+for+Biomedical+Text%2E+Advances+in+Inf)

*Updated 24 October 2026*