Here, the quantitative theory of translation is shown to be of great utility in describing scientific networks. In fact, we
deduce a new Zipf's Law for the descriptors of a set of documents, based on the concepts of centres of interest and of irreversible
parallel translations. This new law can be generalized to other phenomena, such as the distribution of the sizes of cocitation
clusters. Finally, we have established the model, for descriptor presence in a network, which closely fits the values recorded.