A Lexicon-Based Approach for Detecting Hedges in Informal Text

What it is about

The authors want sentence-level hedge detection in spontaneous talk rather than in scientific prose, using eight translated oral-history interviews. They hand-build three lexicons: 76 hedge words (epistemic verbs, adverbs, adjectives, modals, approximators), hedging discourse markers (well, I mean, in a sense, anyway…) and boosters (absolutely, clearly, sure…). A sentence is hedged if it contains a hedge word that survives a dependency-based disambiguation rule, a phrase close to a discourse marker (Jaccard similarity of word sets at about 0.8), or a booster preceded by «not» or «without» («not sure»). On 3,000 sentences annotated by three people the hedge class gets P 32 / R 88 / F1 47 against the majority label and F1 61 against "any annotator said hedge": high recall, low precision, because the cue words often do other jobs.

Ideas

• I1. Port the per-verb disambiguation rules to UD trees in prag/src/feats.rs: think, believe, guess, hope, doubt count as hedges only as root with a first-person nsubj; think followed by a preposition («think about») is not; suppose with an xcomp marked to («supposed to») is not; tend, appear only with an xcomp/ccomp; likely as amod of a noun is not; rather before than is not; should in should have is not. Today hedge fires on string match of prag/data/cues.tsv; measure hedge macro-F1 in prag eval (cross-validated) before and after.
• I2. Add the negated-booster rule: a booster (sure, certain, definitely, clearly) under negation (infer::negated) is a hedge («I'm not sure»). Add a boost class to cues.tsv so the same list later feeds a booster label, which chapter 32 lists as missing.
• I3. The low precision warns that a cue list alone over-fires in dialogue: count, on the 2,986 silver sentences, how often each hedge cue co-occurs with silver hedge=1, and drop or condition cues below a precision floor (a per-cue table written into chapter 32 §9.2).

For Mova

32. Politeness and hedging §7 and §9–10 (hedging devices, the hedge label, computational hedge detection); prag/src/feats.rs (hedge slot), prag/data/cues.tsv; measure: hedge accuracy and macro-F1 in prag eval, plus a negative control (rules on shuffled sentences).

Topics: Discourse analysis

Cite

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

@inproceedings{islam2020lexicon,
  title = {{A Lexicon-Based Approach for Detecting Hedges in Informal Text}},
  author = {Islam, Jumayel and Xiao, Lu and Mercer, Robert E.},
  booktitle = {Proceedings of the Twelfth Language Resources and Evaluation Conference},
  year = {2020},
  url = {https://aclanthology.org/2020.lrec-1.380/}
}

References

19 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Agarwal, S. and Yu, H. (2010). Detecting hedge cues and their scope in biomedical text with conditional random fields. Journal of Biomedical Informatics, 43(6):953– 961. find
2. Ahn, J. J. (2010). Exile as forced migrations: A sociological, literary, and theological approach on the displacement and resettlement of the Southern Kingdom of Judah, volume 417. Walter de Gruyter. find
3. Crystal, D. (1988). On keeping one’s hedges in order. English Today, 4(3):46–47. find
4. Farkas, R., Vincze, V., Móra, G., Csirik, J., and Szarvas, G. (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—Shared Task, pages 1–12. Association for Computational Linguistics. find
5. Ganter, V. and Strube, M. (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. Association for Computational Linguistics. find
6. He, A. W. (1993). Exploring modality in institutional interactions: Cases from academic counselling encounters. Text-Interdisciplinary Journal for the Study of Discourse, 13(4):503–528. find
7. Holmes, J. (1984). Modifying illocutionary force. Journal of Pragmatics, 8(3):345–365. find
8. Jucker, A. H. (1993). The discourse marker well: A relevance-theoretical account. Journal of Pragmatics, 19(5):435–452. find
9. Layman, L. (2009). Reticence in oral history interviews. The Oral History Review, 36(2):207–230. find
10. Light, M., Qiu, X. Y., and Srinivasan, P. (2004). The language of bioscience: Facts, speculations, and statements in between. In HLT-NAACL 2004 Workshop: Linking Biological Literature, Ontologies and Databases. find
11. Manning, C. D., Surdeanu, M., Bauer, J., Finkel, J., Bethard, S. J., and McClosky, D. (2014). The Stanford CoreNLP natural language processing toolkit. In Association for Computational Linguistics (ACL) System Demonstrations, pages 55–60. find
12. Medlock, B. and Briscoe, T. (2007). Weakly supervised learning for hedge classification in scientific literature. In Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, pages 992–999. find
13. Özgür, A. and Radev, D. R. (2009). Detecting speculations and their scopes in scientific text. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 3-Volume 3, pages 1398–1407. Association for Computational Linguistics. find
14. Ponterotto, D. (2018). Hedging in political interviewing. Pragmatics and Society, 9(2):175–207. find
15. Schiffrin, D. (1987). Discourse markers (Studies in Interactional Sociolinguistics, 5). Cambridge: Cambridge University Press. find
16. Szarvas, G., Vincze, V., Farkas, R., and Csirik, J. (2008). The bioscope corpus: annotation for negation, uncertainty and their scope in biomedical texts. In Proceedings of the Workshop on Current Trends in Biomedical Natural Language Processing, pages 38–45. Association for Computational Linguistics. find
17. Szarvas, G. (2008). Hedge classification in biomedical texts with a weakly supervised selection of keywords. Proceedings of ACL-08: HLT, pages 281–289. find
18. Theil, C. K., Stajner, S., and Stuckenschmidt, H. (2018). Word embeddings-based uncertainty detection in financial disclosures. In Proceedings of the First Workshop on Economics and Natural Language Processing, pages 32–37. find
19. Ulinski, M., Benjamin, S., and Hirschberg, J. (2018). Using hedge detection to improve committed belief tagging. In Proceedings of the Workshop on Computational Semantics beyond Events and Roles, pages 1–5. find our card

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/islam-2020-lexicon-based-approach-detecting-hedges/