The following considerations were taken into account concerning the content of our experimental Adaptive Web network.
1. The content of hypertext networks strongly biases and limits its structure. Followingly, our choice for the content of our Adaptive Web network had to be as a-selective as possible.
2. The size of our network had to be large enough to allow for extensive, deep browsing, without having to return upon ones previous path. Yet, the size of the network had to remain manageable for implementation and data-analysis.
We opted for the 150 most frequent English nouns, as derived from the LOB corpus [Johansson & Hofland, 1989) (a frequency analysis of a large base of spoken and written English). As far as semantics and associative relations are concerned, word-frequency seemed to be relatively a-select criteria for the selection of our words. Also, frequent nouns are also often those that have the best generally understood meaning.