SOMLib: A Distributed Digital Library System based on Self-Organizing Maps

We describe an architecture for a distributed digital library system based on an unsupervised neural network model, namely the Self-Organizing Map. The system allows the clustering of text documents forming the basis for intelligent information retrieval. User prooles can be combined with full text queries or sample texts to locate documents within the library system. Contrary to conventional approaches consisting of one central library, the distributed SOMLib allows the combination and integration of several independent SOM-based library systems in a not necessarily hierarchical order. Furthermore, the system is not limited to a speciic type or setup of the basic SOM architecture. Rather, variations of the basic architecture which might serve special needs of some users can be combined.