Combining databases and knowledge bases for assisted browsing
Kristian J. Hammond, Robin Burke · National Conference on Artificial Intelligence · 1995
Finding items of interest in a large multidimensional information space is a problem of increasing importance given the ever-increasing amount of information that is accessible on-line. Standard approaches such as keyword retrieval demand more specificity than the average user can supply. The user must know what he or she wants well enough to create a well-defined query in a query language. The alternative to querying is browsing. Browsing allows users who cannot specify exactly what they seek to rummage around in an information space to find it. Browsing in a large space presents additional problems, however. It is difficult to structure a browsing space so that users can move about in useful and efficient ways without getting lost and without having artificial restrictions on their means of access. Both querying and browsing assume that there is a single location where information is located, but in many cases, multiple sources of information may be relevant and these sources may need to be gathered and combined to meet a particular need. Our approach to problems of information finding in large multi-dimensional information spaces is to employ assisted browsing. The user is provided a standard browsing interface to move about in an information space, but his or her progress in this space is monitored and relevant assistance is provided. The aim of assisted browsing is to allow access to information along a multitude of dimensions and from a multitude of sources without the user needing to be aware of these complexities. Since browsing is the central metaphor, we avoid forcing users to create a specific queries. At the same time, the intelligent assistance available in the system has the ability to draw in other sources of knowledge. Knowledge-based retrieval agents are aware of all of the dimensions of the information and present suggestions that lead the user’s search in reasonable directions. We have drawn our inspiration in this work from case-based reasoning theories of cognition [6,8,9]. Assisted browsing uses the cycle of retrieve and adapt that is fundamental to the case-based reasoning model. We also employ knowledge-based metrics of relevance and similarity that are at the core of many case retrieval systems [2,4].