2003 ImprovingWSDinLexicalChaining

From GM-RKB
Jump to navigation Jump to search

Subject Headings: Lexical Chaining Algorithm, Lexical Chaining Task, Supervised Word Sense Disambiguation Algorithm, Word Sense Disambiguation Algorithm, LexChainer.

Notes

Cited By

~78 http://scholar.google.com/scholar?cites=17419641442926847082

Quotes

Abstract

1. Introduction

  • Passages from spoken or written text have a quality of unity that arises in part from the surface properties of the text; syntactic and lexical devices can be used to create a sense of connectedness between sentences, a phenomenon known as textual cohesion [Halliday and Hasan, 1976]. Of all cohesion devices, lexical cohesion is probably the most amenable to automatic identification [Hoey, 1991]. Lexical cohesion arises when words are related semantically, for example in reiteration relations between a term and a synonym or superordinate.
  • Lexical chaining is the process of connecting semantically related words, creating a set of chains that represent different threads of cohesion through the text. This intermediate representation of text has been used in many natural language processing applications, including automatic summarization [Barzilay and Elhadad, 1997; Silber and McCoy, 2003], information retrieval [Al-Halimi and Kazman, 1998], intelligent spell checking [Hirst and St-Onge, 1998], topic segmentation [Kan et al., 1998], and hypertext construction [Green, 1998].
  • A first computational model of lexical chains was introduced by Hirst and St-Onge [1998]. This linear-time algorithm, however, suffers from inaccurate WSD, since their greedy strategy immediately disambiguates a word as it is first encountered. Later research [Barzilay and Elhadad, 1997] significantly alleviated this problem at the cost of a worse running time (quadratic); computational inefficiency is due to their processing of many possible combinations of word senses in the text in order to decide which assignment is the most likely. More recently, Silber and McCoy [2003] presented an efficient linear-time algorithm to compute lexical chains, which models Barzilay’s approach, but nonetheless has inaccuracies in WSD.
  • In this paper, we further investigate the automatic identification of lexical chains for subsequent use as an intermediate representation of text. In the next section, we propose a new algorithm that runs in linear time and adopts the assumption of one sense per discourse [Gale et al., 1992]. We suggest that separating WSD from the actual chaining of words can increase the quality of chains. In the last section, we present an evaluation of the lexical chaining algorithm proposed in this paper, and compare it against [Barzilay and Elhadad, 1997; Silber and McCoy, 2003] for the task of WSD. This evaluation shows that our algorithm performs significantly better than the other two.

4. Conclusions

  • In this paper, we presented an efficient linear-time algorithm to build lexical chains, showing that one sense per discourse can improve performance. We explained how the separation of WSD from the construction of the chains enables a simplification of the task and improves running time. The evaluation of our algorithm against two known lexical chaining algorithms shows that our algorithm is more accurate when it chooses the senses of nouns to include in lexical chains. The implementation of our algorithm is freely available for educational or research purposes at http://www.cs.columbia.edu/~galley/research.html.

References

  • [Al-Halimi and Kazman, 1998] R. Al-Halimi and R. Kazman. Temporal indexing of video through lexical chaining. In WordNet: An electronic lexical database. MIT Press, 1998.
  • [Barzilay and Elhadad, 1997] Regina Barzilay and M. Elhadad. Using lexical chains for text summarization. In: Proceedings of the Intelligent Scalable Text Summarization Workshop (ISTS’97), ACL, 1997.
  • [Barzilay, 1997] Regina Barzilay. Lexical chains for summarization. Master’s thesis, Ben-Gurion University, 1997.
  • [Gale et al., 1992] W. Gale, Kenneth W. Church, and David Yarowsky. One sense per discourse. In: Proceedings of the DARPA Speech and Natural Language Workshop, 1992.
  • [Green, 1998] S. Green. Automated link generation: Can we do better than term repetition? In: Proceedings of the 7th International World-Wide Web Conference, 1998.
  • [Halliday and Hasan, 1976] M. Halliday and R. Hasan. Cohesion in English. Longman, London, 1976.
  • [Hirst and St-Onge, 1998] Graeme Hirst and D. St-Onge. Lexical chains as representations of context for the detection and correction of malapropisms. In WordNet: An electronic lexical database. MIT Press, 1998.
  • [Hoey, 1991] M. Hoey. Patterns of lexis in text. Oxford University Press, 1991.
  • [Kan et al., 1998] M.-Y. Kan, J. Klavans, and Kathleen R. McKeown. Linear segmentation and segment significance. In: Proceedings of of the 6th Workshop on Very Large Corpora (WVLC-98), 1998.
  • [Miller, 1990] G. Miller. WordNet: An on-line lexical database. International Journal of Lexicography, 3(4):235–312, 1990.
  • [Silber and McCoy, 2003] G. Silber and K. McCoy. Efficiently computed lexical chains as an intermediate representation for automatic text summarization. Computational Linguistics, 29(1), 2003

BibTeX

@inproceedings{Galley&McKeown:03, author = "Galley, Michel and Kathleen R. McKeown", title = "Improving Word Sense Disambiguation in Lexical Chaining", booktitle = "Proceedings of 18th International Joint Conference on Artificial Intelligence ({IJCAI--03})", url = "http://www.cs.columbia.edu/nlp/papers/2003/galley_mckeown_03.pdf", pages = "1486--1488", year = 2003 } ,


 AuthorvolumeDate ValuetitletypejournaltitleUrldoinoteyear
2003 ImprovingWSDinLexicalChainingKathleen R. McKeown
Michel Galley		Improving Word Sense Disambiguation in Lexical Chaininghttp://www.cs.columbia.edu/nlp/papers/2003/galley mckeown 03.pdf
Retrieved from "http://www.gabormelli.com/RKB/index.php?title=2003_ImprovingWSDinLexicalChaining&oldid=689535"