Document Clustering Using the 1 + 1 Dimensional Self-Organising Map

Automatic clustering of documents is a task that has become increasingly important with the explosion of online information. The Self Organising Map (SOM) has been used to cluster documents effectively, but efforts to date have used a single or a series of 2-dimensional maps. Ideally, the output of a document-clustering algorithm should be easy for a user to interpret. This paper describes a method of clustering documents using a series of 1-dimensional SOM arranged hierarchically to provide an intuitive tree structure representing document clusters. Wordnet is used to find the base forms of words and only cluster on words that can be nouns.