Haystack ’ s User Experience for Interacting with Semistructured Information

A major stumbling block in the use of information management tools such as desktops, email managers, file browsers and web browsers is the mismatch between the user’s mental model of information objects and the system’s implementation model. While a user may look at a name on the screen and think of a person, the system may consider it dead text. A user who wants to remember that a certain email message was sent by a distant cousin may be foiled by the fact that email management and family relationship records are the domains of two separate applications that do not talk to each other. The goal of our Haystack system is to reduce the mismatch between the user’s and the system’s models of information, creating a user-object-oriented interface. To achieve this goal, we model the user’s data uniformly and at fine levels of granularity using the Resource Description Framework (RDF). We demonstrate how this internal system model of data can be reflected out to the user interface so that the user can manipulate UI representations of the information in a uniform manner that makes sense to the user. We explain how RDF itself can be used to build a UI framework that performs such reflection systematically. Furthermore, we propose that operations on and mechanisms for organizing RDF data be modeled in RDF as well, making the whole system reflexive and generic. We envision that the user’s everyday use of our system for interacting with the information that he or she cares about will generate metadata and contribute to the global pool of RDF data. The Semantic Web can readily leverage the information collected by Haystack that matters to the end users.