Elamur Research⌕
☰
home / papers / 2014 / An Account of Opinion Implicatures

An Account of Opinion Implicatures

Paper · 2014 · Janyce Wiebe and Lingjia Deng

Summary

Most sentiment analysis work handles opinions that are stated outright. It ignores attitudes that a reader infers by default. Wiebe and Deng propose a rule-based framework for one family of these opinion implicatures: those that arise when explicit sentiment co-occurs with events that help or harm an entity (goodFor/badFor, or "gfbf", events).

To keep errors from automatic components out of the reasoning, the system takes manually annotated sentiments and events as input. It stores its knowledge in a graph and chains defeasible rule schemas forward inside the nested belief and sentiment spaces of different holders. A chain halts when contrary evidence appears, for example a sign that an act was not volitional. From a sentence like John was glad that Mary saved Bill it derives that John feels positively toward Bill and toward Mary, and that John believes Mary favours Bill.

The report is a conceptual "what-if" blueprint rather than a benchmarked system. Companion studies show that using these rules as constraints for sentiment propagation helps automatic systems. This matters because it moves opinion analysis from surface polarity toward reasoning about who feels what toward whom.

Key points

  • The input is gold annotations of explicit sentiments and gfbf events. The output is a set of implicit sentiments and beliefs, often of mixed polarity, attributed to nested sources such as (writer, Japan, Mr. Bush). The full system output for the worked examples is in Appendix A.
  • The system has ten rule schemas, a mechanism for inference within belief and sentiment spaces, and explicit cases where inference is blocked. Influencer chains that end in gfbf events are composed before inference begins, and Rule 10 relies on word connotation.
  • Deng and Wiebe (2014) encode implicature constraints in a graph and run Loopy Belief Propagation, which improves over an explicit-sentiment classifier.
  • A separate ILP study combines four local classifiers through global inference. It gives substantial gains on two of the ambiguities and no loss on the others.
  • Both automatic studies are limited. They cover only the writer's sentiments and use simplified versions of four of the ten schemas.
  • The architecture draws on several earlier lines of work:
    • abduction (Hobbs et al.);
    • default belief ascription (Wilks & Ballim);
    • script-style forward inference (Schank's SAM);
    • multiple belief spaces (SNePS, Martins).
  • The volition-based blocking follows Greene & Resnik's finding that more volitional harmful acts draw harsher judgments of the agent.

Interesting ideas

  • I1. Polarity can be propagated through events with simple sign rules. If an event harms something a holder dislikes, the holder likes the event, and the agent of a liked event is liked too; this yields attitudes toward entities with no sentiment words at all.
  • I2. Sources are treated as attribution chains, with each holder owning a separate belief/sentiment space. By default a holder is assumed to share the outer view unless there is evidence to the contrary.
  • I3. Inferences are defaults that run forward until contrary evidence breaks the chain, such as an action that was not volitional.
  • I4. The reasoning layer is evaluated on gold annotations, which separates its quality from upstream noise.
  • I5. The implicature rules double as constraints for joint inference, shrinking the space of combined interpretations instead of committing to a pipeline.

For Mova

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

  • I1 (frames): prag could attach goodFor/badFor labels to VerbNet-based frames and derive implicit attitudes toward agents and patients, logging each rule application as an explainable step.
  • I2 (world): the world of a text could hold per-holder sub-worlds keyed by attribution chains, so that the system can answer who believes or likes what without merging conflicting viewpoints.
  • I3 (frames): implicature rules in prag could carry explicit blocking conditions (intentionality, accident markers), so that withdrawn inferences are recorded and explainable rather than silently absent.
  • I4 (learn): the frames and world stages could be tested on hand-annotated input alongside parser output, so that the gap between the two runs measures how much error comes from upstream.
  • I5 (learn): disagreements between witness parsers on roles or sentiment could be resolved by an ILP over candidates that respect implicature consistency.

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

🎓University of PittsburghPittsburgh · United States38 works · 21,591 citations

📚 Shelves 5 Computer Science · 9,086 works Artificial Intelligence · 8,889 works Arts and Humanities · 9,296 works Language and Linguistics · 9,206 works Literature and Literary Theory · 340 works

🧭 Topics 2 Sentiment & emotion · 608 works Implicature · 72 works

Cite

Cite as Janyce Wiebe and Lingjia Deng. 2014. An Account of Opinion Implicatures. arXiv:1404.6491.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1404.6491 · arXiv 1404.6491

@misc{wiebe2014account,
  title = {{An Account of Opinion Implicatures}},
  author = {Janyce Wiebe and Lingjia Deng},
  year = {2014},
  doi = {10.48550/arxiv.1404.6491},
  eprint = {1404.6491},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1404.6491}
}

References

45 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Anand P, Reschke K (2010) Verb classes as evaluativity functor classes. In: Interdisciplinary Workshop on Verbs. The Identification and Representation of Verb Features Banfield A (1982) Unspeakable Sentences. find
  2. Routledge and Kegan Paul, Boston Breck E, Choi Y, Cardie C (2007) Identifying expressions of opinion in context. In: IJCAI’07: Proceedings of the 20th international joint conference on Artifical intelligence, Morgan Kaufmann Publishers Inc., San Francisco, CA, USA, pp 2683–2688 find
  3. Bruder G, Wiebe J (1995) Recognizing subjectivity and identifying subjective characters in third-person fictional narrative. In: Duchan J, Bruder G, Hewitt L (eds) Deixis in Narrative: A Cognitive Science Perspective, Lawrence Erlbaum Associates Choi Y, Cardie C (2008) Learning with compositional semantics as structural inference for subsentential sentiment analysis. In: Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics, Honolulu, Hawaii, pp 793–801, URL http://www.aclweb.org/anthology/D08-1083 ↗ our card
  4. Choi Y, Cardie C, Riloff E, Patwardhan S (2005) Identifying sources of opinions with conditional random fields and extraction patterns. In: HLT ’05: Proceedings of the conference on Human Language Technology and Empirical Methods in Natural Language Processing, Association for Computational Linguistics, Morristown, NJ, USA, pp 355–362 our card
  5. Choi Y, Breck E, Cardie C (2006) Joint extraction of entities and relations for opinion recognition. In: Proceedings of EMNLP Das S, Chen M (2001) Yahoo for amazon: Extracting market sentiment from stock message boards. In: Proceedings of the 8th Asia Pacific Finance Association Annual Conference. find
  6. Dave K, Lawrence S, Pennock DM (2003) Mining the peanut gallery: opinion extraction and semantic classification of product reviews. In: WWW ’03: Proceedings of the 12th international conference on World Wide Web, ACM, New York, NY, USA, pp 519–528 find
  7. Deng L, Wiebe J (2014) Sentiment propagation via implicature constraints. In: Meeting of the European Chapter of the Association for Computational Linguistics (EACL-2014) our card
  8. Deng L, Choi Y, Wiebe J (2013) Benefactive/malefactive event and writer attitude annotation. In: Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), Association for Computational Linguistics, Sofia, Bulgaria, pp 120–125, URL http://www.aclweb.org/anthology/P13-2022 ↗
  9. Doran R, Ward G, Larson M, McNabb Y, Baker RE (2012) A novel experimental paradigm for distinguishing between ‘what is said’ and ‘what is implicated’. Language 88(1):124–154 find
  10. Feng S, Bose R, Choi Y (2011) Learning general connotation of words using graph-based algorithms. In: Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics, Edinburgh, Scotland, UK., pp 1092–1103, URL http://www.aclweb.org/anthology/D11-1101 ↗
  11. Feng S, Kang JS, Kuznetsova P, Choi Y (2013) Connotation lexicon: A dash of sentiment beneath the surface meaning. In: Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), Association for Computational Linguistics, Sofia, Bulgaria, pp 1774–1784, URL http://www.aclweb.org/anthology/P13-1174 ↗
  12. Goyal A, Riloff E, Daumé III H (2010) Automatically producing plot unit representations for narrative text. In: Empirical Methods in Natural Language Processing (EMNLP), Boston, MA, URL http://pub.hal3.name/#daume10daal ↗
  13. Goyal A, Riloff E, Daume III H (2012) A computational model for plot units. ↗
  14. Computational Intelligence Greene S, Resnik P (2009) More than words: Syntactic packaging and implicit sentiment. In: Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics, Association for Computational Linguistics, Boulder, Colorado, pp 503–511, URL http://www.aclweb.org/anthology/N/N09/N09-1057 ↗ our card
  15. Grice HP (1967) Logic and conversation. The William James lectures Grice HP (1989) Studies in the Way of Words. find
  16. Harvard University Press Hobbs JR, Stickel ME, Appelt DE, Martin P (1993) Interpretation as abduction. Artificial Intelligence 63(1-2):69–142 find
  17. Kim SM, Hovy E (2005) Automatic detection of opinion bearing words and sentences. In: Companion Volume to the Proceedings of IJCNLP-05, the Second International Joint Conference on Natural Language Processing, Jeju Island, KR, pp 61–66 find
  18. Lehnert WG (1981) Plot Units and Narrative Summarization. Cognitive Science 5(4):293–331 ↗
  19. Levinson S (1983) Pragmatics. Cambridge Textbooks in Linguistics, Cambridge University Press, URL http://books.google.com/books?id=SJXr9w\_lVLUC ↗
  20. Liu B (2012) Sentiment Analysis and Opinion Mining. ↗
  21. Morgan & Claypool Moilanen K, Pulman S (2007) Sentiment composition. In: Proceedings of RANLP 2007, Borovets, Bulgaria Moilanen K, Pulman S (2009) Multi-entity sentiment scoring. In: Proceedings of the International Conference RANLP-2009, Association for Computational Linguistics, Borovets, Bulgaria, pp 258–263, URL http://www.aclweb.org/anthology/R09-1048 ↗
  22. Moilanen K, Pulman S, Zhang Y (2010) Packed feelings and ordered sentiments: Sentiment parsing with quasi-compositional polarity sequencing and compression. In: Proceedings of the 1st Workshop on Computational Approaches to Subjectivity and Sentiment Analysis (WASSA 2010), pp 36–43 find
  23. Neviarouskaya A, Prendinger H, Ishizuka M (2011) Affect analysis model: novel rule-based approach to affect sensing from text. Natural Language Engineering 17(01):95–135 ↗
  24. Pang B, Lee L (2008) Opinion mining and sentiment analysis. Found Trends Inf Retr 2(1-2):1–135 ↗
  25. Pang B, Lee L, Vaithyanathan S (2002) Thumbs up?: sentiment classification using machine learning techniques. In: EMNLP ’02: Proceedings of the ACL-02 conference on Empirical methods in natural language processing, Association for Computational Linguistics, Morristown, NJ, USA, pp 79–86 ↗ our card
  26. Pearl J (1982) Reverend bayes on inference engines: A distributed hierarchical approach. In: Proceedings of the American Association of Artificial Intelligence National Conference on AI, Pittsburgh, PA, pp 133–136 find
  27. Popescu AM, Etzioni O (2005) Extracting product features and opinions from reviews. In: Proceedings of the Human Language Technologies Conference/Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP-2005), Vancouver, Canada, pp 339–346, URL http://www.aclweb.org/anthology/H/H05/H05-1043.pdf Qiu G, Liu B, Bu J, Chen C (2011) Opinion word expansion and target extraction through double propagation. Comput Linguist 37:9–27 ↗ our card
  28. Quirk R, Greenbaum S, Leech G, Svartvik J (1985) A Comprehensive Grammar of the English Language. find
  29. Longman, New York Rapaport WJ (1986) Logical foundations for belief representation. Cognitive Science 10(4):371–422 ↗
  30. Reschke K, Anand P (2011) Extracting contextual evaluativity. In: Ninth International Conference on Computational Semantics (IWCS 2011) find
  31. Riloff E, Qadir A, Surve P, De Silva L, Gilbert N, Huang R (2013) Sarcasm as contrast between a positive sentiment and negative situation. In: Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, Association for Computational Linguistics, Seattle, Washington, USA, pp 704– 714, URL http://www.aclweb.org/anthology/D13-1066 ↗ our card
  32. Ruppenhofer J, Rehbein I (2012) Semantic frames as an anchor representation for sentiment analysis. In: Proceedings of the 3rd Workshop in Computational Approaches to Subjectivity and Sentiment Analysis, Association for Computational Linguistics, Jeju, Korea, pp 104–109, URL http://www.aclweb.org/anthology/W12-3716 ↗
  33. Schank R, Abelson R (1977) Scripts, Plans, Goals and Understanding. find
  34. Lawrence Erlbaum Associates, Hillsdale, New Jersey Shapiro SC, Rapaport WJ (1987) Sneps considered as a fully intensional propositional semantic network. In: Cercone N, McCalla G (eds) The Knowledge Frontier, Springer, New York, pp 262–315 find
  35. Stein D, Wright S (eds) (1995) Subjectivity and Subjectivisation. find
  36. Cambridge University Press, Cambridge Stoyanov V, Cardie C (2008) Topic identification for fine-grained opinion analysis. In: Proceedings of COLING Wiebe J (1994) Tracking point of view in narrative. Computational Linguistics 20(2):233–287 our card
  37. Wiebe J, Mihalcea R (2006) Word sense and subjectivity. In: Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, Association for Computational Linguistics, Sydney, Australia, pp 1065–1072, DOI 10.3115/1220175.1220309, URL http://www.aclweb.org/anthology/P06-1134.pdf ↗
  38. Wiebe J, Riloff E (2005) Creating subjective and objective sentence classifiers from unannotated texts. In: Sixth International Conference on Intelligent Text Processing and Computational Linguistics (CICLing-2005), Mexico City, Mexico Wiebe J, Wilson T, Cardie C (2005) Annotating expressions of opinions and emotions in language. Language Resources and Evaluation (formerly Computers and the Humanities) 39(2/3):164–210 find
  39. Wilensky R (1978) Understanding goal-based stories. PhD thesis, Yale University, Also available as Tech Report YALEU/CSD/RR #140 find
  40. Wilks Y, Ballim A (1987) Multiple agents and the heuristic ascription of belief. In: Proceedings of the 10th International Joint Conference on Artificial Intelligence - Volume 1, IJCAI’87, pp 118–124 find
  41. Wilks Y, Bien J (1983) Beliefs, points of view and multiple environments. Cognitive Science 7:95–119 ↗
  42. Wilson T, Wiebe J, Hoffmann P (2005) Recognizing contextual polarity in phrase-level sentiment analysis. In: Proceedings of the Human Language Technologies Conference/Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP-2005), Vancouver, Canada, pp 347–354 our card
  43. Yi J, Nasukawa T, Bunescu R, Niblack W (2003) Sentiment analyzer: Extracting sentiments about a given topic using natural language processing techniques. Data Mining, IEEE International Conference on 0:427 find
  44. Zhang L, Liu B (2011) Identifying noun product features that imply opinions. In: Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, Association for Computational Linguistics, Portland, Oregon, USA, pp 575–580, URL http://www.aclweb.org/anthology/P11-2101 ↗
  45. Zúñiga F, Kittilä S (2010) Introduction. In: Zúñiga F, Kittilä S (eds) Benefactives and malefactives, Typological studies in language, J. Benjamins Publishing Company, URL http://books.google.com/books?id=B8tzvbjuRfwC ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wiebe-2014-account-opinion-implicatures/