Efficiently Computed Lexical Chains as an Intermediate Representation for Automatic Text Summarization

Paper · 2002 · H. Gregory Silber and Kathleen F. McCoy

Summary

Lexical chains are groups of semantically related nouns linked by identity, synonymy, hypernymy/hyponymy or sibling relations. Barzilay and Elhadad (1997) proposed them as an intermediate representation for summarization. Earlier ways of building them were either manual or exponential in cost, so they could not be applied to texts of realistic length. Silber and McCoy reformulate the full Barzilay–Elhadad chaining method so that it runs in linear time and space relative to the number of nouns.

Rather than listing every possible interpretation, the system fills an array of "metachains" indexed by WordNet sense, which encodes all interpretations implicitly. A second pass then assigns each noun to the chain it strengthens most, scored by relation type and textual distance. The authors give a complexity proof and propose a way to evaluate this intermediate stage alone. Chains are computed for both source documents and their human-written summaries, and the concepts the two share are compared. On 24 documents, the outcome was judged promising. The work matters because it makes chain-based summarization practical for large document collections.

Key points

Resources and preprocessing. The system relies on WordNet 1.6, with the noun database recompiled into a memory-mapped binary array for fast lookup. A part-of-speech tagger identifies the nouns. The metachain array is sized as the number of WordNet noun senses plus the number of document nouns, so that words missing from WordNet can still be handled.
Empirically tuned scores. Identical words and synonyms always contribute 1. Hypernyms contribute 1 within one sentence and 0.5 beyond that. Siblings contribute 1, 0.3 or 0.2 as distance grows, and 0 by default. Ties go to the higher (more specific) WordNet sense number.
Strong chains. Following Barzilay and Elhadad, strong chains are those scoring more than two standard deviations above the document's mean chain score.
Complexity. Running time grows linearly, O(n), with the count of nouns in the text. WordNet's size is treated as a constant in the proof: 94,474 nouns and 66,025 synsets in version 1.6.
Limitations and future work. Proper nouns, missing anaphora resolution and gaps in WordNet coverage remain problems; terms absent from WordNet fall back to frequency counts. Any comparable lexicon could replace WordNet. The authors plan to feed the chains into a text generation step, and an online demo and search engine are available.

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 DelawareNewark ·

United States20 works · 1,056 citations

Fields, topics and concepts

📚 Shelves 4Computer Science · 16,667 works Artificial Intelligence · 16,194 works Arts and Humanities · 16,039 works Language and Linguistics · 15,891 works

🧭 Topics 1Summarization · 232 works

Cite

Cite as H. Gregory Silber and Kathleen F. McCoy. 2002. Efficiently Computed Lexical Chains as an Intermediate Representation for Automatic Text Summarization. Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1162/089120102762671954

@article{silber2002efficiently,
  title = {{Efficiently Computed Lexical Chains as an Intermediate Representation for Automatic Text Summarization}},
  author = {H. Gregory Silber and Kathleen F. McCoy},
  journal = {Computational Linguistics},
  year = {2002},
  doi = {10.1162/089120102762671954},
  url = {https://doi.org/10.1162/089120102762671954}
}

References

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

Barzilay, Regina and Michael Elhadad. 1997. Using lexical chains for text summarization. In Proceedings of the Intelligent Scalable Text Summarization Workshop (ISTS-97), Madrid, Spain.
Barzilay, Regina, Kathleen R. McKeown, and Michael Elhadad. 1999. Information fusion in the context of multi-document summarization. In Proceedings of the 37th Annual Conference of the Association for Computational Linguistics, College Park, MD. Association for Computational Linguistics, New Brunswi…
Hirst, Graeme and David St.-Onge. 1997. Lexical chains as representation of context for the detection and correction of malapropisms. In Christiane Fellbaum, editor, Wordnet: An electronic lexical database and some of its applications. MIT Press, Cambridge, pages 305–332. find
Jing, H. and K. McKeown. 2000. Cut and paste based text summarization. In Proceedings of NAACL-00, Seattle. find
Mani, Inderjeet, Barbara Gates, and Eric Bloedorn. 1999. Improving summaries by revising them. In Proceedings of the 37th Annual Conference of the Association for Computational Linguistics, College Park, MD. Association for Computational Linguistics, New Brunswick, NJ. find
Marcu, Daniel. 1999. The automatic creation of large scale corpora for summarization research. In The 22nd International ACM SIGIR Conference on Research and Development in Information Retrieval, Berkeley. ACM Press, New York. find
Morris, J. and G. Hirst. 1991. Lexical cohesion computed by thesaural relations as an indecator of the structure of text. Computational Linguistics, 18(1):21–45. find
Silber, H. Gregory and Kathleen F. McCoy. 2000. Efficient text summarization using lexical chains. In 2000 International Conference on Intelligent User Interfaces, New Orleans, January. find
Sparck Jones, Karen. 1993. What might be in summary? Information Retrieval ’93, Regensburg, Germany, September, pages 9–26. find

Updated 30 October 2026

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/silber-2002-efficiently-computed-lexical-chains-as/