Building an Adaptive Hierarchy of Clusters for Text Data

Text clustering has been recognized as an important component in Web-based applications. Clustering data on a hierarchical structure enables exploring data on different levels of granularity, providing a more intuitive view that is close to the way humans view the world. Self-Organizing Map (SOM) based models have been found to have certain advantages for clustering sizeable text data. However, current existing approaches lack in providing an adaptive hierarchical structure within in a single model. This paper proposes an unsupervised hierarchical clustering approach based on the Growing Self-Organizing Map (GSOM). By utilizing GSOM’s spread factor, our approach offers an adaptive architecture with the capability of detecting necessary layers to form a hierarchy, avoiding a number of issues that a traditional top-down or bottom-up hierarchical clustering approach often encounter. Experiment has shown that this approach has the potential for efficiently clustering heterogeneous text data.