Concept taxonomy is an integral part of a classification system, however, it is not sufficient for instantiating a knowledge base to train an automated classifier. The complete knowledge base requires an ontology, a set of instances, and functions that instantiate concepts and relations. Accordingly, the document classifier, a system for classifying concepts (topics) contained in text documents, may be trained from a properly organized collection of topic descriptions. Given a repository of diverse documents, the problem is how to translate human objectives into an organized set of exemplars. In this paper, we describe a document ontology discovery tool for constructing knowledge bases. The user interacts with a repository indexed by latent semantic indexing and models the ontology presented as a hierarchy of topics. Automatically generated concept taxonomies produce results not only comparable with other clustering techniques but in addition, provide meaningful topic descriptions necessary for human-computer interaction in ontology modeling.
[1]
Steffen Staab,et al.
Ontology Learning for the Semantic Web
,
2002,
IEEE Intell. Syst..
[2]
David M. Pennock,et al.
Inferring hierarchical descriptions
,
2002,
CIKM '02.
[3]
T. Landauer,et al.
Indexing by Latent Semantic Analysis
,
1990
.
[4]
David D. Lewis,et al.
Reuters-21578 Text Categorization Test Collection, Distribution 1.0
,
1997
.
[5]
Pavel Berkhin,et al.
A Survey of Clustering Data Mining Techniques
,
2006,
Grouping Multidimensional Data.
[6]
Lyle H. Ungar,et al.
Statistical Relational Learning for Link Prediction
,
2003
.
[7]
George Karypis,et al.
Evaluation of hierarchical clustering algorithms for document datasets
,
2002,
CIKM '02.
[8]
J. W. Semich.
Science Applications International Corp.
,
1993
.