Exploration of Document Collections with Self-Organizing Maps: A Novel Approach to Similarity Representation

摘要: Classification is one of the central issues in any system dealing with text data. The need for effective approaches dramatically increased nowadays due to advent massive digital libraries containing free-form documents. What we are looking powerful methods exploration such whereby detection similarities between various documents overall goal. In other words, that may be used gain insight inherent structure items contained a archive needed. this paper demonstrate applicability self-organizing maps, neural network model adhering unsupervised learning paradigm, task document clustering. order improve representation result present an extension basic rule captures movement weight vectors two-dimensional output space convenient visual inspection. extended training algorithm allows intuitive analysis input data and most important, recognition cluster boundaries.