Taxonomy-based Document Clustering
Masoud Makrehchi · Journal of Digital Information Management · 2011
One well-known document representation for text clus- tering is bag-of-words. Although it is simple and popular, it ignores semantics, underly ing linguistic information, and word correlations. In this paper, Bag-Of-Queries, a new document representation is proposed. First, a taxonomy of the terms in the local dictionary derived for data set is extracted. Ex tracting taxonomy is performed by learning term dependencies using an information theoretic in- clusion index. Next, the taxonomy is partitioned to generate a set of correlated terms or bag of queries. Since every two partitions of the taxonomy belong to two different concepts, they are con sidered semantically orthogonal queries. This provides a new space of or- thogonal features, which is necessary for an effective clustering. As a result, instead of using terms as features, they are employed to build a set of queries. Documents are ranked in response to the queries using a similarity measure such as Cosine. The similarity indices are consid ered as new features in a vector space model representation. The proposed approach outperforms bag of word based document representation for clus tering. It also extracts new non-redundant features and at the same time reduces dimensionality.