Title: Dimensionality reduction for blog tag mining

Authors: Flora S. Tsai

Addresses: School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore 639798, Singapore

Abstract: Blog tags are labels of blog documents that classify them into different categories. Most tags are user-generated, which create problems such as inconsistencies in tags across different users, blogs without tags, lack of descriptive tags, lack of semantic distinction, etc. In this paper, we utilise dimensionality reduction techniques to reduce the inherent noise in blog tags. A tag-topic model is combined with dimensionality reduction, and then evaluated on real-world blog data. By employing dimensionality reduction techniques to reduce the document-tag space, better classification results were achieved. This indicates that the noise in tags can be effectively reduced by representing the original set of tags with a smaller number of latent tags, which can lead to more accurate real-time categorisation of blog documents.

Keywords: weblogs; blog tags; dimensionality reduction; autoencoders; tag topics; categorisation; blogs; blog tag mining; text mining.

DOI: 10.1504/IJWET.2011.040726

International Journal of Web Engineering and Technology, 2011 Vol.6 No.3, pp.286 - 298

Published online: 28 Feb 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article