The unceasing growth of electronic information available on the Web has made it indispensable to develop tools to analyse and evaluate its quality. Given the subjectivity underlying most of the qualitative analytical indicators at the present time, we here propose the use of characteristics or indicators of a formal character. We apply Kohonen's neural networks and study their topological organization in order to analyse how Web spaces behave with respect to these networks' formal characteristics. The interpretation of the results brings out the underlying structures and relationships in a closed Web environment.