SorTables: a browser for a digital library

Much research in information retrieval has focused more on matching results to queries than on browsing those results. After brie y exploring browsing in physical and electronic libraries, we introduce SortTables, a new system that focuses on support for browsing. We explore the evolution of the system in light of early implementation experience and formative evaluation of the interface. Finally, we brie y review related work, and discuss future directions. 1 Browsing in the Physical Library In a physical library, we think of browsing as moving among the shelves, looking at items of possible interest. Several characteristics of this type of browsing stand out: access to the actual items, potential access to all the items, a sense of neighborhood, variable focus, and the opportunity for serendipity. Physical browsing provides access to the actual items. For example, books provide powerful cues about their meaning in a situated way, through physical cues as well as content. A worn-out book might be regarded as more (or less) interesting because of its apparent popularity. A thick book might look like too much work; one with an interesting cover might be chosen instead. Browsing potentially provides access to all the items. One has the sense that one could start at one end and go to the other, looking at each book. This is di erent from trying to retrieve books by generating queries: we don't have to know all the \query terms" in advance, and we don't have to deal with the same item twice (because it was retrieved by di erent queries). The sense of neighborhood is an important reason that browsing is e ective. Since books are organized by call number, and call numbers are assigned in a way that keeps related books together, we often nd books on related topics close together. Browsing allows a variable focus: in a promising area, one can systematically examine all items; when items in that area aren't useful, one can make a sweeping movement To appear in CIKM '95, The 4th International Conference on Information and Knowledge Management, Nov. 28{Dec. 2, 1995, Baltimore, Maryland. Draft: August 14, 1995. through large areas, skipping the uninteresting ones. Finally, browsing allows for a natural serendipity: one can choose a book at random from a particular area, or from a random area of the library. As with variable focus, one has control over how random one's selection can be. 2 Browsing in the Electronic Library Browsing in the electronic library can retain many of these characteristics, except for access to the physical items. Electronic browsing can provide something not feasible in the physical library: multiple orderings of the items. We're not restricted to ordering strictly by call number: we can think of browsing by author or date. (One could imagine a library where the three-dimensional arrangement has significance; for example, the oor of an item could correspond to year of publication or publication type. However, the particulars of the three-dimensional arrangement are usually only accidental, as shelves are usually linearly organized by call number.) Before computers, card catalogs o ered some of this facility: there were often card catalogs organized by author, subject, and title. It would be unusual, however, to nd a card catalog devoted to organizing items by publication date or date of acquisition. Electronic browsing can overcome some of the limitations of the paper card catalogs. Electronic browsers can be much faster to use (as a patron can sit at a terminal, perhaps in his own o ce, rather than moving around the physical catalog). Browsers can provide greater exibility in manipulating the items. For example, it is possible to search on both authors and years, and to present the items in a variety of ways (e.g., sorted by publication date). 3 SortTables: A System for Browsing SortTables is an interface metaphor developed to support browsing. Information retrieval systems have often lacked a browser, thus forgoing the bene ts of this type of interaction, or they have provided a simple electronic realization of the card catalog, making little use of the manipulative capabilities of the computer. SortTables is not in any sense a complete solution to the problems of digital libraries. It only addresses the problem of presenting and manipulating a set of items; as presented here, its search facilities are limited. It doesn't address issues of how documents are stored or retrieved, protocols, security,