A system for organizing a content site so that articles preferred by a user (viewer) of the site are brought to the fore for easy access. The system observes the user's actions during the normal course of browsing through a content site, and creates a model of the user's preferences for various types of articles. This model is created as an Internet user `clicks` on articles which the user desires to read, without requiring any other feedback from the user. The user model is then employed to reorganize the content site so that the articles preferred by the user are presented in an order according to the user's interests. This model can also be used to present the user with advertising material based on the user's demonstrated interests. The system performs the above functions by using word vector-space representation of the documents combined with adaptive learning techniques. A word vector for a document is created by counting all the occurrences of each word in a document and creating a vector whose components comprise the word frequencies. A document is represented by a point in a high-dimensional space whose axes represent the words in a given dictionary. Thus, similar documents are close together in this vector-space. The word vector of an article forms the input to an adaptive ranking engine. The output of the ranking engine is a value which represents the strength of a particular user's preference for reading that article. In this manner, the contents of an online newspaper or an archive of any type can be rank ordered by the numerical value of the output of the ranking system.