Elamur Research⌕
☰
home / papers / 2014 / Sentiment Propagation via Implicature Constraints

Sentiment Propagation via Implicature Constraints

Summary

Most sentiment analysis systems only detect opinions that are stated outright. Writers often reveal attitudes indirectly, through events that help or harm entities. Deng and Wiebe study whether such implicit opinions can be recovered by reasoning over goodFor/badFor (gfbf) events.

They define four defeasible implicature rule schemas, with sixteen rules in total. The rules connect the writer's sentiment toward a gfbf event with the sentiment toward its agent and its object. These constraints become edge potentials in a graph where nodes are entities (agents and objects). Explicit sentiment scores serve as node priors, and Loopy Belief Propagation spreads polarity across the graph. Influencers that reverse or retain an event's effect are folded into badFor and goodFor edges.

The graph is built from manually annotated gfbf information in a corpus of editorials and blogs about the Affordable Care Act. Compared with an explicit-only sentiment classifier, the model raises precision by roughly 10 points and also improves recall. Seeded with one correct node per subgraph, it labels the remaining nodes correctly about 89% of the time.

The work shows that sentiment inference through event structure can substantially strengthen fine-grained sentiment analysis.

Key points

  • Data: The corpus comes from Deng et al. (2013) and is available via mpqa.cs.pitt.edu.
    • It contains 134 documents with 1,762 annotated gfbf triples.
    • Of these, 692 are goodFor/retainer and 1,070 are badFor/reverser.
    • Writer attitudes toward noun phrases are mostly polar: 1,495 positive, 1,114 negative and only 8 neutral.
  • Rules and influencers: The four rule schemas cover inference in both directions, from event to object or agent and back. Reversers (e.g., prohibit) are treated as badFor events and retainers as goodFor events.
  • Inference: Loopy Belief Propagation runs for at most 100 iterations, and convergence took about 34 iterations on average. The explicit-sentiment baseline barely beats the Majority and GFBF baselines, which reach about 56% precision.
  • Graph-only evaluation: In this setup, each node in turn is given its gold label and the others are predicted from it.
    • Across 983 subgraphs, correctness is 0.8874.
    • For the 169 subgraphs with more than two nodes, it rises to 0.9030.
  • Limitations:
    • Gold gfbf annotations were used to build the graph, so this is not yet an end-to-end system.
    • About 41.6% of errors come from nodes that never received a polar value.
    • The agent rules (schema 3) can fail when an action is accidental or forced.

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: Implicature

Cited by

34 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 32 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/v1/e14-1040

@inproceedings{deng2014sentiment,
  title = {{Sentiment Propagation via Implicature Constraints}},
  author = {Deng, Lingjia and Wiebe, Janyce},
  booktitle = {Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics},
  year = {2014},
  doi = {10.3115/v1/e14-1040},
  url = {https://doi.org/10.3115/v1/e14-1040}
}

References

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

  1. Pranav Anand and Kevin Reschke. 2010. Verb classes as evaluativity functor classes. In Interdisciplinary Workshop on Verbs. The Identification and Representation of Verb Features. find
  2. Yejin Choi and Claire Cardie. 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, pages 793–801, Honolulu, Hawaii, October. Association for Computational Linguistics. find
  3. Pradeep Dasigi, Weiwei Guo, and Mona Diab. 2012. Genre independent subgroup detection in online discussion threads: A study of implicit attitude using textual latent semantics. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 65–69, Jeju Island, Korea, July. Association for Computational Linguistics. find
  4. Lingjia Deng, Yoonjung Choi, and Janyce Wiebe. 2013. Benefactive and malefactive event and writer attitude annotation. In 51st Annual Meeting of the Association for Computational Linguistics (ACL-2013, short paper). find
  5. Song Feng, Jun Sak Kang, Polina Kuznetsova, and Yejin Choi. 2013. Connotation lexicon: A dash of sentiment beneath the surface meaning. In Proceedings of the 51th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), Sofia, Bulgaria, Angust. Association for Computational Linguistics. find
  6. Amit Goyal, Ellen Riloff, and Hal Daum III. 2012. A computational model for plot units. Computational Intelligence, pages 466–488. ↗
  7. Rodney D. Huddleston and Geoffrey K. Pullum. 2002. The Cambridge Grammar of the English Language. Cambridge University Press, April. ↗ our card
  8. Richard Johansson and Alessandro Moschitti. 2013. Relational features in fine-grained opinion analysis. Computational Linguistics, 39(3). ↗
  9. Bing Liu. 2012. Sentiment Analysis and Opinion Mining. Morgan & Claypool. ↗
  10. Karo Moilanen and Stephen Pulman. 2007. Sentiment composition. In Proceedings of RANLP 2007, Borovets, Bulgaria. find
  11. Karo Moilanen, Stephen Pulman, and Yue Zhang. 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), pages 36–43. find
  12. J. Pearl. 1982. Reverend bayes on inference engines: A distributed hierarchical approach. In Proceedings of the American Association of Artificial Intelligence National Conference on AI, pages 133–136, Pittsburgh, PA. find
  13. Kevin Reschke and Pranav Anand. 2011. Extracting contextual evaluativity. In Proceedings of the Ninth International Conference on Computational Semantics, IWCS ’11, pages 370–374, Stroudsburg, PA, USA. Association for Computational Linguistics. find
  14. Ellen Riloff, Ashequl Qadir, Prafulla Surve, Lalindra De Silva, Nathan Gilbert, and Ruihong Huang. 2013. Sarcasm as contrast between a positive sentiment and negative situation. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 704–714, Seattle, Washington, USA, October. Association for Computational Linguistics. find our card
  15. P.J. Stone, D.C. Dunphy, M.S. Smith, and D.M. Ogilvie. 1966. The General Inquirer: A Computer Approach to Content Analysis. MIT Press, Cambridge. find
  16. Chenhao Tan, Lillian Lee, Jie Tang, Long Jiang, Ming Zhou, and Ping Li. 2011. User-level sentiment analysis incorporating social networks. In Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining, pages 1397–1405. ACM. find
  17. Xiaolong Wang, Furu Wei, Xiaohua Liu, Ming zhou, and Ming Zhang. 2011. Topic sentiment anaylsis in twitter: A graph-based hashtag sentiment classification appraoch. In CIKM, pages 1031–1040. find
  18. Janyce Wiebe, Theresa Wilson, and Claire Cardie. 2005. Annotating expressions of opinions and emotions in language ann. Language Resources and Evaluation, 39(2/3):164–210. ↗
  19. Theresa Wilson, Janyce Wiebe, and Paul Hoffmann. 2005. Recognizing contextual polarity in phrase-level sentiment analysis. In HLT/EMNLP, pages 347–354. find
  20. Jonathan S Yedidia, William T Freeman, and Yair Weiss. 2005. Constructing free-energy approximations and generalized belief propagation algorithms. Information Theory, IEEE Transactions on, 51(7):2282–2312. ↗
  21. Lei Zhang and Bing Liu. 2011. Identifying noun product features that imply opinions. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 575–580, Portland, Oregon, USA, June. Association for Computational Linguistics. find
  22. F. Zúñiga and S. Kittilä. 2010. Introduction. In F. Zúñiga and S. Kittilä, editors, Benefactives and malefactives, Typological studies in language. J. Benjamins Publishing Company. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/deng-2014-sentiment-propagation-implicature-constraints/