Department of Software Technology
Vienna University of Technology
Finding Structure in Text Archives
Abstract:
With the advance and massive growth of electronic text
archives, the need for tools emerges, which help to gain
insight into the basic structure of the underlying digital
library. We present a neural network approach for the
analysis and exploration of text archives aiming at the
detection and visualization of the inherent structure of
the text collection. This cluster visualization technique
called Adaptive Coordinates is based on an extended
learning rule for the self-organizing map. It provides
an intuitive visualization by mapping clusters in a
high-dimensional input space onto groups of nodes in a
2-dimensional output space. We further compare the results
of this mapping with another prominent cluster visualization
technique, namely Sammon's Mapping.
Up
Comments: rauber@ifs.tuwien.ac.at