Elamur Research⌕
☰
home / papers / 2014 / Abstractive Summarization of Product Reviews Using Discourse Structure

Abstractive Summarization of Product Reviews Using Discourse Structure

Summary

The paper tackles abstractive opinion summarization of many product reviews. Earlier systems either stitched together extracted sentences, depended on hand-built feature taxonomies, or needed training data to decide aspect order. The authors start from reviews whose aspects and polarity/strength (P/S) scores (integers from -3 to +3) are already annotated. Each review is parsed with an RST discourse parser (Joty et al., 2013), and every leaf is reduced to the aspect words it contains, giving an Aspect-based Discourse Tree (ADT). The ADTs for one product are merged into an Aggregated Rhetorical Relation Graph (ARRG). Weighted PageRank ranks the aspects, and the top-ranked subgraph becomes an Aspect Hierarchy Tree (AHT). A product-independent, template-based NLG step then writes text that reports how opinions are distributed and how aspects explain one another. Two crowdsourced user studies on reviews of twelve products found the system significantly preferred over extractive and abstractive baselines. The work shows that discourse structure can drive both content selection and summary organization without any domain knowledge.

Key points

  • Pipeline: RST discourse trees → ADTs (aspect-only leaves; aspect-less EDUs kept and marked "no aspect") → ARRG → Weighted PageRank subgraph selection → AHT → template-based micro planning and sentence realization.
  • Aspect extraction and sentiment are taken as given input. Aspect coreference (e.g., "g3", "canon g3", "canon" → "camera") was merged by hand to reduce noise.
  • Baselines included the extractive MEAD* and LR and the abstractive SA. Raters preferred the proposed system over all of them in two user studies, one without the source reviews and one with them, covering twelve products.
  • Inter-rater agreement was 65% in study 1 and 70% in study 2, where raters also saw the reviews. Agreement was measured on 100 randomly sampled crowdsourcing units.
  • In the qualitative feedback, raters praised the grammar, aspect coverage, and the coherent organization of related features. Some still preferred the technical detail found in extractive summaries.

Interesting ideas

  • I1. Strip discourse trees down to the entities of interest. Keeping the RST skeleton but replacing each EDU with only its aspects turns a general parse into a task-specific structure. The same trick could apply to other entity types, such as places or traded goods, in other genres.
  • I2. Aggregate noisy per-document structures into a single graph. Merging many imperfect discourse trees into one weighted relation graph smooths out parser errors on informal text. This is helpful wherever individual parses are unreliable but the collection is large.
  • I3. Rank with PageRank over rhetorical links instead of a trained ordering model. Using graph centrality on discourse relations to pick and order content needs no training data. That makes it attractive for low-resource languages or new domains.
  • I4. Use templates that state opinion distribution and causal links between aspects. Phrases like "X% of reviewers… mainly because of Y" give quantitative, explanatory output without deep NLG machinery. This pattern is reusable in any aggregated-feedback report.
  • I5. Run the evaluation with and without the source documents. Comparing rater judgments in both settings shows how much access to the inputs changes preferences and agreement. This is a cheap way to probe how robust a human evaluation is.

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: Natural language processing

Topics Discourse analysis Summarization

 Institutions

🎓University of British ColumbiaVancouver · Canada34 works · 32,862 citations🎓Università della Svizzera italianaLugano · Switzerland2 works · 639 citations

Cite

Cite as Shima Gerani, Yashar Mehdad, Giuseppe Carenini, Raymond T. Ng and Bita Nejat. 2014. Abstractive Summarization of Product Reviews Using Discourse Structure. Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/v1/d14-1168

@inproceedings{gerani2014abstractive,
  title = {{Abstractive Summarization of Product Reviews Using Discourse Structure}},
  author = {Shima Gerani, Yashar Mehdad, Giuseppe Carenini, Raymond T. Ng and Bita Nejat},
  booktitle = {Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
  year = {2014},
  doi = {10.3115/v1/d14-1168},
  url = {https://doi.org/10.3115/v1/d14-1168}
}

References

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

  1. Dan Ariely, George Loewenstein, and Drazen Prelec. 2003. “Coherent Arbitrariness”: Stable Demand Curves Without Stable Preferences. Quarterly Journal of Economics, 118:73–105. find
  2. Nicholas Asher, Farah Benamara, and Yvette Yannick Mathieu. 2008. Distilling opinion in discourse: A preliminary study. In Coling 2008: Companion volume: Posters and Demonstrations, pages 5–8. find
  3. Regina Barzilay, Kathleen R. McKeown, and Michael Elhadad. 1999. Information fusion in the context of multi-document summarization. In Proceedings of the 37th annual meeting of the Association for Computational Linguistics on Computational Linguistics, ACL ’99, pages 550–557, Stroudsburg, PA, USA. A… our card
  4. Giuseppe Carenini and Jackie Chi Kit Cheung. 2008. Extractive vs. nlg-based abstractive summarization of evaluative text: the effect of corpus controversiality. In INLG ’08: Proceedings of the Fifth International Natural Language Generation Conference, pages 33–41, Morristown, NJ, USA. Association f… find
  5. Giuseppe Carenini, Jackie Chi Kit Cheung, and Adam Pauls. 2013. Multi-document summarization of evaluative text. Computational Intelligence, 29(4):545–576. find
  6. Giuseppe Di Fabbrizio, Amanda Stent, and Robert Gaizauskas. 2014. A hybrid approach to multi-document summarization of opinions in reviews. In Proceedings of the 8th International Natural Language Generation conference, INLG 2014. find
  7. Günes Erkan and Dragomir R. Radev. 2004. Lexrank: graph-based lexical centrality as salience in text summarization. J. Artif. Int. Res., 22(1):457–479, December. find
  8. Kavita Ganesan, ChengXiang Zhai, and Jiawei Han. 2010. Opinosis: a graph-based approach to abstractive summarization of highly redundant opinions. In Proceedings of the 23rd International Conference on Computational Linguistics, COLING ’10, pages 340–348, Stroudsburg, PA, USA. Association for Comput… our card
  9. Minqing Hu and Bing Liu. 2004a. Mining and summarizing customer reviews. In Proceedings of ACM SIGKDD Conference on Knowledge Discovery and Data Mining 2004 (KDD 2004), pages 168–177, Seattle, Washington. find
  10. Minqing Hu and Bing Liu. 2004b. Mining opinion features in customer reviews. In Proceedings of the Nineteenth National Conference on Artificial Intelligence (AAAI-2004). find
  11. Minqing Hu and Bing Liu. 2006. Opinion feature extraction using class sequential rules. In Proceedings of AAAI 2006 Spring Sympoia on Computational Approaches to Analyzing Weblogs (AAAI-CAAW 2006). find
  12. Shafiq Joty, Giuseppe Carenini, Raymond Ng, and Yashar Mehdad. 2013. Combining intra- and multi-sentential rhetorical parsing for document-level discourse analysis. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 486–496, Sofi… our card
  13. Hyun Duk Kim, Kavita Ganesan, Parikshit Sondhi, and ChengXiang Zhai. 2011. Comprehensive review of opinion summarization. find
  14. Hyun Duk Kim, Malu Castellanos, Meichun Hsu, ChengXiang Zhai, Umeshwar Dayal, and Riddhiman Ghosh. 2013. Compact explanatory opinion summarization. In Proceedings of the 22Nd ACM International Conference on Conference on Information &\#38; Knowledge Management, CIKM ’13, pages 1697–1702, New York, NY… find
  15. Angeliki Lazaridou, Ivan Titov, and Caroline Sporleder. 2013. A bayesian model for joint unsupervised induction of sentiment, aspect and discourse representations. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1630–1639, Sof… find
  16. Kevin Lerman, Sasha Blair-Goldensohn, and Ryan Mc-Donald. 2009. Sentiment summarization: Evaluating and learning user preferences. In Proceedings of the 12th Conference of the European Chapter of the Association for Computational Linguistics, EACL ’09, pages 514–522, Stroudsburg, PA, USA. Associatio… find
  17. Hugo Liu and Push Singh. 2004. Conceptnet: A practical commonsense reasoning toolkit. BT Technology Journal, 22(4):211–226. find
  18. Yue Lu, ChengXiang Zhai, and Neel Sundaresan. 2009. Rated aspect summarization of short comments. In Proceedings of the 18th International Conference on World Wide Web, WWW ’09, pages 131–140, New York, NY, USA. ACM. find
  19. William C. Mann and Sandra A. Thompson. 1988. Rhetorical structure theory: Toward a functional theory of text organization. Text, 8(3):243–281. our card
  20. Subhabrata Mukherjee and Sachindra Joshi. 2013. Sentiment aggregation using conceptnet ontology. In Proceedings of the 6th International Joint Conference on Natural Language Processing, IJCNLP 2013. find
  21. Bo Pang and Lillian Lee. 2004. A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts. In Proceedings of the 42Nd Annual Meeting on Association for Computational Linguistics, ACL ’04, Stroudsburg, PA, USA. Association for Computational Linguistics. our card
  22. Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan. 2002. Thumbs up?: Sentiment classification using machine learning techniques. In Proceedings of the ACL-02 Conference on Empirical Methods in Natural Language Processing - Volume 10, EMNLP ’02, pages 79–86, Stroudsburg, PA, USA. Association for Com… our card
  23. Dragomir Radev, Timothy Allison, Sasha Blair-Goldensohn, John Blitzer, Arda Çelebi, Stanko Dimitrov, Elliott Drabek, Ali Hakim, Wai Lam, Danyu Liu, Jahna Otterbacher, Hong Qi, Horacio Saggion, Simone Teufel, Michael Topper, Adam Winkel, and Zhu Zhang. 2004. MEAD — A platform for multidocument multi… our card
  24. Ehud Reiter and Robert Dale. 2000. Building Natural Language Generation Systems. Cambridge University Press, New York, NY, USA. find
  25. Benjamin Snyder and Regina Barzilay. 2007. Multiple aspect ranking using the good grief algorithm. In HLT-NAACL, pages 300–307. find
  26. Swapna Somasundaran, Galileo Namata, Janyce Wiebe, and Lise Getoor. 2009. Supervised and unsupervised methods in employing discourse relations for improving opinion polarity classification. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing: Volume 1 - Volume 1… our card
  27. Ivan Titov and Ryan T. McDonald. 2008. A joint model of text and aspect ratings for sentiment summarization. In Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics, June 15-20, 2008, Columbus, Ohio, USA, ACL 2008, pages 308–316. Association for Computational Lingu… our card
  28. Rakshit S. Trivedi and Jacob Eisenstein. 2013. Discourse connectors for latent subjectivity in sentiment analysis. In HLT-NAACL, pages 808–813. The Association for Computational Linguistics. find
  29. Wenpu Xing and Ali Ghorbani. 2004. Weighted pagerank algorithm. In Proceedings of the Second Annual Conference on Communication Networks and Services Research, CNSR ’04, pages 305–314. IEEE Computer Society. find
  30. Alexander Yeh. 2000. More accurate tests for the statistical significance of result differences. In Proceedings of the 18th conference on Computational linguistics - Volume 2, COLING ’00, pages 947– 953, Stroudsburg, PA, USA. Association for Computational Linguistics. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/gerani-2014-abstractive-summarization-product-reviews-discourse/