Elamur Research⌕
☰
home / papers / 2009 / More than Words: Syntactic Packaging and Implicit Sentiment

More than Words: Syntactic Packaging and Implicit Sentiment

Paper · 2009 · Stephan Greene and Philip Resnik

Summary

The paper tackles implicit sentiment, or perspective: attitude carried by text that looks neutral and uses almost the same words as an alternative description. The authors argue that this attitude often lives in syntactic "packaging", such as who appears as the subject or whether a verb is nominalized or passivized. They link these choices to 13 grammatically relevant semantic properties drawn from Dowty's proto-roles and from Hopper and Thompson's semantic transitivity, including volition, causation, change of state and telicity. Two psycholinguistic studies built around verbs of killing show that syntactic form shifts readers' sympathy toward a perpetrator and that the semantic properties predict these sentiment ratings. The authors then introduce OPUS (observable proxies for underlying semantics), automatically computed features that approximate these properties. In a supervised SVM setup, OPUS features beat Lin et al.'s earlier results on the Bitter Lemons corpus. The work matters because it ties lexical-semantic theory to perceived sentiment and gives practical features for detecting bias in non-evaluative text.

Key points

  • The authors combine Dowty (1991) with Hopper and Thompson (1980) to get 13 semantic properties. These are grouped by the subject X, the event, and the object Y of a canonical transitive clause.
  • Property-rating study:
    • 18 native English speakers rated descriptions on a 1–7 scale.
    • The descriptions used 11 killing verbs, split into a "transitive" class (e.g. kill, assassinate) and an "ergative" class (e.g. strangle, drown).
    • Each verb appeared in either a transitive form or a nominalized form.
  • Sentiment study:
    • 31 participants rated how sympathetic headlines were to the perpetrator.
    • Headlines came in transitive, nominalized or passive form.
    • Syntactic form had a significant effect (F(2, 369) = 33.902, p < .001).
    • Transitive headlines drew lower sympathy than the other two forms.
  • Bitter Lemons corpus, Test Scenario 1:
    • The corpus has 297 Israeli and 297 Palestinian documents.
    • Across 423 parameter settings, OPUS averaged 95.41% accuracy, with a best of 97.64%.
    • Lin et al.'s baselines were 88.22% (SVM) and 93.46% (NB-B).
    • OPUS fell slightly below NB-B in only 8 runs.
  • Bitter Lemons corpus, Test Scenario 2 (training on editor documents, testing on guest documents):
    • OPUS averaged 83.12%, with a maximum of 85.86%.
    • It fairly consistently beat Lin's SVM (81.48%) and came close to or matched NB-B (85.85%).

Interesting ideas

  • I1. Test the theory with people first. The authors checked their linguistic hypothesis with controlled human ratings before building features. This gives an engineered feature set a principled justification, which helps any project that needs features beyond word lists.
  • I2. Minimal-pair stimuli. The experiments keep the event and nearly all the vocabulary fixed while changing only the syntax. This cleanly separates structural effects from lexical ones and could be reused to probe models or annotators for framing sensitivity.
  • I3. Pair words with their grammatical role. OPUS-style features tie a specific word to its syntactic role rather than using generic structural templates. This applies to bias or stance detection in any language with a parser, including Ukrainian news text.
  • I4. Split data by source, not by document. The evaluation separates training and test data by author type (editors vs. guests) or by site. This ensures the classifier learns viewpoint rather than author style or topic, a useful safeguard against leakage in stance and perspective tasks.
  • I5. Sweep parameters and report the distribution. The authors report results across hundreds of parameter settings (423 runs) instead of a single best run. This shows how robust the gain over baselines is and makes for a more honest comparison.

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

📚 Shelves 6 Computer Science · 9,086 works Artificial Intelligence · 8,889 works Arts and Humanities · 9,296 works Language and Linguistics · 9,206 works Psychology · 798 works Experimental and Cognitive Psychology · 463 works

🧭 Topics 1 Sentiment & emotion · 608 works

Cite

Cite as Stephan Greene and Philip Resnik. 2009. More than Words: Syntactic Packaging and Implicit Sentiment. Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics.

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

@inproceedings{greene2009more,
  title = {{More than Words: Syntactic Packaging and Implicit Sentiment}},
  author = {Stephan Greene and Philip Resnik},
  booktitle = {Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics},
  year = {2009},
  url = {https://aclanthology.org/N09-1057/}
}

References

33 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. John Broder. 2007. Familiar fallback for officials: ’mistakes were made’. New York Times. March 14. find
  2. F. J. Damerau. 1993. Generating and evaluating domain-oriented multi-word terms from texts. Information Processing and Management, 29:433–447. find
  3. Hoa Trang Dang, Karin Kipper, Martha Palmer, and Joseph Rosenzweig. 1998. Investigating Regular Sense Extensions Based on Intersective Levin Classes. In ACL/COLING 98, pages 293–299, Montreal, Canada, August 10–14. find
  4. Bonnie J. Dorr. 1993. Machine Translation: A View from the Lexicon. The MIT Press, Cambridge, MA. find
  5. David Dowty. 1991. Thematic Proto-Roles and Argument Selection. Language, 67:547–619. find
  6. S. D. Durbin, J. N. Richter, and D. Warner. 2003. A system for affective rating of texts. In Proc. 3rd Workshop on Operational Text Classification, KDD-2003. find
  7. Robert M. Entman. 1993. Framing: Toward clarification of a fractured paradigm. Journal of Communication, 43(4):51–58. find
  8. Anthony Fader, Dragomir R. Radev, Michael H. Crespin, Burt L. Monroe, Kevin M. Quinn, and Michael Colaresi. 2007. MavenRank: Identifying influential members of the US Senate using lexical centrality. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP). find
  9. Michael Gamon. 2004. Sentiment classification on customer feedback data: noisy data, large feature vectors, and the role of linguistic analysis. In Proc. COLING. find
  10. M. Gentzkow and J. Shapiro. 2006a. Media bias and reputation. Journal of Political Economy, 114:280– 316. find
  11. M. Gentzkow and J. Shapiro. 2006b. What drives media slant? Evidence from U.S. newspapers. http://ssrn.com/abstract=947640. ↗
  12. Natalie Glance, Matthew Hurst, Kamal Nigam, Matthew Siegler, Robert Stockton, and Takashi Tomokiyo. 2005. Deriving marketing intelligence from online discussion. In Proc. KDD’05, pages 419–428, New York, NY, USA. ACM. find
  13. Stephan Greene. 2007. Spin: Lexical Semantics, Transitivity, and the Identification of Implicit Sentiment. Ph.D. thesis, University of Maryland. find
  14. T. Groseclose and J. Milyo. 2005. A measure of media bias. The Quarterly Journal of Economics, 120:1191– 1237. find
  15. B. Harden. 2006. On Puget Sound, It’s Orca vs. Inc. The Washington Post. July 26, page A3. find
  16. Marti Hearst. 1992. Direction-based text interpretation as an information access refinement. In Paul Jacobs, editor, Text-Based Intelligent Systems, pages 257–274. Lawrence Erlbaum Associates. find
  17. Paul Hopper and Sandra Thompson. 1980. Transitivity in Grammar and Discourse. Language, 56:251–295. find
  18. E. Kako. 2006. Thematic role properties of subjects and objects. Cognition, 101(1):1–42, August. find
  19. Michael Laver, Kenneth Benoit, and John Garry. 2003. Extracting policy positions from political texts using words as data. American Political Science Review, 97(2):311–331. find
  20. M. Lemmens. 1998. Lexical perspectives on transitivity and ergativity. John Benjamins. find
  21. Beth Levin and Malka Rappaport Hovav. 2005. Argument Realization. Research Surveys in Linguistics. Cambridge University Press, New York. find
  22. Beth Levin. 1993. English Verb Classes and Alternations: A Preliminary Investigation. University of Chicago Press, Chicago, IL. find
  23. Wei-Hao Lin, Theresa Wilson, Janyce Wiebe, and Alexander Hauptmann. 2006. Which side are you on? identifying perspectives at the document and sentence levels. In Proceedings of the Conference on Natural Language Learning (CoNLL). find
  24. Lanny W. Martin and Georg Vanberg. 2008. A robust transformation procedure for interpreting political text. Political Analysis, 16(1):93–100. find
  25. G. McKoon and T. MacFarland. 2000. Externally and internally caused change of state verbs. Language, pages 833–858. find
  26. M. Mulder, A. Nijholt, M. den Uyl, and P. Terpstra. 2004. A lexical grammatical implementation of affect. In Proc. TSD-04, Lecture notes in computer science 3206, pages 171–178). Springer-Verlag. find
  27. Tony Mullen and Robert Malouf. 2008. Taking sides: User classification for informal online political discourse. Internet Research, 18:177–190. find
  28. Mari Broman Olsen and Philip Resnik. 1997. Implicit Object Constructions and the (In)transitivity Continuum. In 33rd Proceedings of the Chicago Linguistic Society, pages 327–336. find
  29. Bo Pang and Lillian Lee. 2008. Opinion mining and sentiment analysis. Foundations and Trends in Information Retrieval, 2(1-2):1–135. find
  30. James Pustejovsky. 1991. The Generative Lexicon. Computational Linguistics, 17(4):409–441. find
  31. Philip J. Stone. 1966. The General Inquirer: A Computer Approach to Content Analysis. The MIT Press. find
  32. Matt Thomas, Bo Pang, and Lillian Lee. 2006. Get out the vote: Determining support or opposition from Congressional floor-debate transcripts. In Proc. EMNLP, pages 327–335. our card
  33. Zhibao Wu and Martha Palmer. 1994. Verb Semantics and Lexical Selection. In Proc. ACL, pages 133–138, Las Cruces, New Mexico. our card

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/greene-2009-more-than-words-syntactic-packaging/