Advanced search
1 file | 252.02 KB

Text clustering based on concept-relational decomposition

(2010) ICL 2010 Proceedings. p.357-359
Author
Organization
Abstract
In this paper, a new approach on text clustering is proposed. Based on the concept-relational decomposition, a set of documents can be represented as a multirelation on a concept space C. An interesting property of this representation is that high cuts of the multirelation reveal highly relevant couples of concepts. It appears that each maximal component of a cut corresponds to a cluster on the condition that intra-component and inter-component dependencies are taken into account. The optimal cut is the cut for which the number of clusters best fits the estimated number of clusters. An experiment conducted on 550 Dutch documents shows that our method outperforms standard clustering algorithms that operate on a vector space.

Downloads

  • (...).pdf
    • full text
    • |
    • UGent only
    • |
    • PDF
    • |
    • 252.02 KB

Citation

Please use this url to cite or link to this publication:

Chicago
Bronselaer, Antoon, Saskia Debergh, Dirk Van Hyfte, and Guy De Tré. 2010. “Text Clustering Based on Concept-relational Decomposition.” In ICL 2010 Proceedings, 357–359. Vienna, Austria: International Association on Online Engineering.
APA
Bronselaer, A., Debergh, S., Van Hyfte, D., & De Tré, G. (2010). Text clustering based on concept-relational decomposition. ICL 2010 Proceedings (pp. 357–359). Presented at the Interactive Computer Aided Learning (ICL 2010), Vienna, Austria: International Association on Online Engineering.
Vancouver
1.
Bronselaer A, Debergh S, Van Hyfte D, De Tré G. Text clustering based on concept-relational decomposition. ICL 2010 Proceedings. Vienna, Austria: International Association on Online Engineering; 2010. p. 357–9.
MLA
Bronselaer, Antoon, Saskia Debergh, Dirk Van Hyfte, et al. “Text Clustering Based on Concept-relational Decomposition.” ICL 2010 Proceedings. Vienna, Austria: International Association on Online Engineering, 2010. 357–359. Print.
@inproceedings{1043293,
  abstract     = {In this paper, a new approach on text clustering is proposed. Based on the concept-relational decomposition, a set of documents can be represented as a multirelation on a concept space C. An interesting property of this representation is that high cuts of the multirelation reveal highly relevant couples of concepts. It appears that each maximal component of a cut corresponds to a cluster on the condition that intra-component and inter-component dependencies are taken into account. The optimal cut is the cut for which the number of clusters best fits the estimated number of clusters. An experiment conducted on 550 Dutch documents shows that our method outperforms standard clustering algorithms that operate on a vector space.},
  author       = {Bronselaer, Antoon and Debergh, Saskia and Van Hyfte, Dirk and De Tr{\'e}, Guy},
  booktitle    = {ICL 2010 Proceedings},
  isbn         = {9783899585414},
  language     = {eng},
  location     = {Hasselt, Belgium},
  pages        = {357--359},
  publisher    = {International Association on Online Engineering},
  title        = {Text clustering based on concept-relational decomposition},
  year         = {2010},
}