Braque: Design of an Interface to Support User Interaction in Information Retrieval

We suggest that information retrieval is most appropriately considered as an inherently interactive process, and describe the design of an interface to a bibliographic information retrieval system that supports user interaction in an integrated fashion. A significant aspect of the interface design is its dependence upon a two-level hypertext model of information retrieval system databases, and the seamless support of a variety of information seeking strategies through the use of this model, and of a model of dimensions of information seeking behaviors. 1. INFORMATION RETRIEVAL AS INTERACTION When searching for information in some knowledge resource such as a textual database, people engage in a variety of behaviours, which we can term information seeking strategies. These include, for example: searching for some known and identifiable item(s);* searching for items similar to some known item; searching for items on some identified topic; looking around for something interesting among items; inspecting items and their contents; identifying useful items by inspection; and browsing among item descriptors and item organization schemes. This list of information seeking strategies (ISSs) is not meant to be exhaustive, but rather indicative of the wide range of such behaviours. We note that these ISSs appear to have some characteristics that they share, yet also serve to distinguish them from one another. Additionally, the observation that, in systems not constrained to a single ISS, people often change from one strategy to others during the course of a single information seeking episode, is of special interest. Finally, we suggest that all of these ISSs can be construed as interactions between the user and other aspects of the IR system. In this paper, we address some implications of these issues for the design of IR systems, and present an IR system interface design based upon them. The ISSs we have identified above seem to be differentiable according to their values on a relatively small set of factors, or dimensions. For instance, the first, looking for a specific known item, is the classic example of a well defined search, whereas looking around for something interesting is prototypical scanning. ISSs like inspecting items and their contents, or browsing among item descriptors, may have the goal of learning about the relevant issues, whereas identifying useful items has the goalof selection. Looking around in a group of items can also be characterized as retrieval by recognition, whereas searching for items on some identified topic is an example of retrieval by specification. And interaction with descriptors or organisation schemes of items can be contrasted with interaction with information items themselves, as a meta-information/information distinction. From these rather informal examples, we can identify four dimensions or factors of ISS: l method of interaction (scanning-searching); l goal of interaction (learning-selecting); l mode of retrieval (recognition-specification); and l resource considered (information items-meta-information). *By item we mean here either an information object, such as a document or an image, or identifiable subunits of such objects.