Modeling Information Content Using Observable Behavior

Abstract

This paper presents a framework for modeling the content of information objects such as documents and video programs based on observation of how users interact with those objects in the course of information seeking and use. Four categories of potentially observable user behaviors are identified: examination, retention, reference, and annotation. The framework draws together techniques from the information filtering, Web searching and citation indexing, and identifies the natural scope (portion of an object, complete object, or collection of objects) at which each behavior can be observed. The process of using observations as a basis for identifying information that may be of interest to specific users is addressed briefly, and alternative system architectures are proposed. The paper concludes by identifying some open issues that could have significant implications for the utility of information content models that are based on observable behavior. 1.