Elamur Research⌕
☰
home / papers / 2008 / Scientific Paper Summarization Using Citation Summary Networks

Scientific Paper Summarization Using Citation Summary Networks

Paper · 2008 · Vahed Qazvinian and Dragomir R. Radev

Summary

The paper addresses the difficulty of entering an unfamiliar research area. It treats summarizing a single article as a first step toward summarizing a whole topic. The key input is the article's citation summary: the set of sentences in other papers that cite it. Each citing sentence tends to highlight a different contribution.

The data come from the ACL Anthology Network (AAN). The authors built five topic clusters (DP, PBMT, Summ, QA, TE) and picked 25 target papers, five per cluster. Annotators listed the contributions ("facts") they found in each citation summary. These annotations were turned into fact distribution matrices and a pyramid-style weighting, where a fact counts more if more citing sentences mention it.

The method builds a lexical similarity network over citing sentences, finds communities in it, and draws summary sentences from separate clusters. Two variants were tested: C-RR and C-lexrank. For 5-sentence summaries, C-lexrank scored highest among the systems (mean pyramid score 0.75), ahead of Lexrank (0.71). The work shows that citation text is a usable, multi-perspective source for summarizing scientific papers.

Key points

  • The AAN covers almost 11,000 ACL papers. Its citation network has 8,898 nodes and 38,765 directed edges.
  • The five clusters were formed by matching a topic phrase in paper titles or content. Their sizes range from 56 nodes (TE) to 839 nodes (Summ). Citation summaries of the 25 evaluated papers range from 8 to 180 sentences.
  • Similarity measures were compared by how many fact-sharing sentence pairs appear among the most similar pairs:
    • tf-idf (General) found 34, 66 and 74 such pairs in the top 100, 200 and 300.
    • Levenshtein found only 2, 9 and 16.
  • Mean pyramid scores for |S| = 5:
    • Gold: 0.99
    • Random: 0.41
    • C-RR: 0.69
    • Lexrank: 0.71
    • C-lexrank: 0.75
  • Facts were kept only if several annotators shared them; facts found by just one annotator were dropped.

Interesting ideas

  • I1. Others' citations as a summary source. Sentences written about a paper by other authors can replace or complement its abstract. They capture how the community actually perceives the paper, which could help with summarizing or indexing any heavily cited document.
  • I2. Choosing a similarity measure by a downstream criterion. Candidate measures were ranked by how often their top-scoring pairs share an annotated fact, not by intuition. This cheap diagnostic fits any graph-based method that needs edges reflecting the same underlying content.
  • I3. One sentence per cluster for coverage. Taking sentences from different communities of the sentence graph pushes the summary to cover distinct contributions rather than repeating the dominant one. The paper also notes a pitfall: plain round-robin selection can pick sentences that carry no fact, because every sentence ends up in some cluster.
  • I4. Frequency-based fact pyramid. A fact's weight equals the number of sentences that mention it, and the score is normalized by the best achievable weight for the summary length. This yields an evaluation that is aware of an upper bound, and it can be reused wherever redundancy across many sources signals importance.
  • I5. Bracketing systems between baselines. A greedy gold summary and a random summary are reported side by side, so each system's score can be read against both the ceiling and the floor.

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 MichiganAnn Arbor · United States36 works · 12,181 citations

📚 Shelves 7 Computer Science · 9,086 works Artificial Intelligence · 8,889 works Arts and Humanities · 9,296 works Language and Linguistics · 9,206 works Decision Sciences · 85 works Statistics, Probability and Uncertainty · 32 works History and Philosophy of Science · 18 works

🧭 Topics 1 Summarization · 159 works

Cite

Cite as Vahed Qazvinian and Dragomir R. Radev. 2008. Scientific Paper Summarization Using Citation Summary Networks. Proceedings of the 22nd International Conference on Computational Linguistics (Coling 2008).

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

@inproceedings{qazvinian2008scientific,
  title = {{Scientific Paper Summarization Using Citation Summary Networks}},
  author = {Vahed Qazvinian and Dragomir R. Radev},
  booktitle = {Proceedings of the 22nd International Conference on Computational Linguistics (Coling 2008)},
  year = {2008},
  doi = {10.3115/1599081.1599168},
  url = {https://doi.org/10.3115/1599081.1599168}
}

References

25 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Barzilay, Regina, Kathleen R. McKeown, and Michael Elhadad. 1999. Information fusion in the context of multi-document summarization. In ACL’99, pages 550–557. our card
  2. Bradshaw, Shannon. 2002. Reference Directed Indexing: Indexing Scientific Literature in the Context of Its Use. Ph.D. thesis, Northwestern University. find
  3. Bradshaw, Shannon. 2003. Reference directed indexing: Redeeming relevance for subject search in citation indexes. In Proceedings of the 7th European Conference on Research and Advanced Technology for Digital Libraries. find
  4. Carbonell, Jaime G. and Jade Goldstein. 1998. The use of MMR, diversity-based reranking for reordering documents and producing summaries. In SIGIR’98, pages 335–336. find
  5. Clauset, Aaron, Mark E. J. Newman, and Cristopher Moore. 2004. Finding community structure in very large networks. Phys. Rev. E, 70(6):066111, Dec. find
  6. Elkiss, Aaron, Siwei Shen, Anthony Fader, Güneş Erkan, David States, and Dragomir R. Radev. 2008. Blind men and elephants: What do citation summaries tell us about a research article? Journal of the American Society for Information Science and Technology, 59(1):51–62. find
  7. Erkan, Güneş and Dragomir R. Radev. 2004. Lexrank: Graph-based centrality as salience in text summarization. Journal of Artificial Intelligence Research (JAIR). find
  8. Erkan, Güneş. 2006. Language model-based document clustering using random walks. In Proceedings of the HLT-NAACL conference, pages 479–486, New York City, USA, June. Association for Computational Linguistics. find
  9. Joseph, Mark T. and Dragomir R. Radev. 2007. Citation analysis, centrality, and the ACL Anthology. Technical Report CSE-TR-535-07, University of Michigan. Department of Electrical Engineering and Computer Science. find
  10. Kan, Min-Yen, Judith L. Klavans, and Kathleen R. McKeown. 2002. Using the Annotated Bibliography as a Resource for Indicative Summarization. In Proceedings of LREC 2002, Las Palmas, Spain. find
  11. Kupiec, Julian, Jan Pedersen, and Francine Chen. 1995. A trainable document summarizer. In SIGIR ’95, pages 68–73, New York, NY, USA. ACM. find
  12. Levenshtein, Vladimir I. 1966. Binary Codes Capable of Correcting Deletions, Insertions and Reversals. Soviet Physics Doklady, 10:707. find
  13. Lin, Jimmy J. and Dina Demner-Fushman. 2006. Methods for automatically evaluating answers to complex questions. Information Retrieval, 9(5):565–587. find
  14. Manning, Christopher D., Prabhakar Raghavan, and Hinrich Schütze. 2008. Introduction to Information Retrieval. Cambridge University Press. find
  15. Marton, Gregory and Alexey Radul. 2006. Nuggeteer: Automatic nugget-based evaluation using descriptions and judgements. In Proceedings of NAACL/HLT. find
  16. Nanba, Hidetsugu and Manabu Okumura. 1999. Towards multi-paper summarization using reference information. In IJCAI1999, pages 926–931. find
  17. Nanba, Hidetsugu, Takeshi Abekawa, Manabu Okumura, and Suguru Saito. 2004a. Bilingual presri: Integration of multiple research paper databases. In Proceedings of RIAO 2004, pages 195–211, Avignon, France. find
  18. Nanba, Hidetsugu, Noriko Kando, and Manabu Okumura. 2004b. Classification of research papers using citation links and citation types: Towards automatic review article generation. In Proceedings of the 11th SIG Classification Research Workshop, pages 117– 134, Chicago, USA. find
  19. Nenkova, Ani and Rebecca Passonneau. 2004. Evaluating content selection in summarization: The pyramid method. Proceedings of the HLT-NAACL conference. find
  20. Newman, Mark E. J. 2001. The structure of scientific collaboration networks. PNAS, 98(2):404–409. find
  21. Radev, Dragomir, 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 multi-document mul… find
  22. Siddharthan, Advaith and Simone Teufel. 2007. Whose idea was this, and why does it matter? attributing scientific work to citations. In Proceedings of NAACL/HLT-07. find
  23. Teufel, Simone and Marc Moens. 2002. Summarizing scientific articles: experiments with relevance and rhetorical status. Comput. Linguist., 28(4):409–445. our card
  24. Teufel, Simone, Advaith Siddharthan, and Dan Tidhar. 2006. Automatic classification of citation function. In Proceedings of the EMNLP, Sydney, Australia, July. our card
  25. Teufel, Simone. 2005. Argumentative Zoning for Improved Citation Indexing. Computing Attitude and Affect in Text: Theory and Applications, pages 159– 170. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/qazvinian-2008-scientific-paper-summarization-citation-summary/