|
|||
Research into context-aware applicationsContext-aware retrieval: a key need for the futureAn aim of many researchers is to identify a key market need that will arise, say, three years in the future, and to perform research and build prototype systems to meet that need. I am convinced that one such market need will be retrieving information by context. There are already commercial systems that retrieve information according to the user's current location ("Find the nearest Macdonalds"). As such systems become more wide-ranging, and use richer contexts and more diverse information sources, there will be an increasing need for retrieval systems that both give good continuous performance and retrieve information that precisely meets the user's current needs. Since moving to the University of Exeter in 2000, I have been working on context-aware retrieval, and have gained greatly from working with Gareth Jones, who is an expert in retrieval. (Gareth has recently moved to Dublin City University, but we continue to work together.) Retrieval has made huge strides, the most obvious of which are manifested in web search engines. Context-aware retrieval needs to make similar strides in order to have the market impact that has been forecast. Our interest is that context-aware retrieval is both tantalisingly similar and tantalising different from conventional Information Retrieval and Information Filtering. Our aim is to understand these differences properly and to build prototypes for the next generation of context-aware retrieval systems. A prototype that includes our latest ideas based on best-match retrieval has been built; it has been implemented by Lindsey Ford, an RA on the project. We are very grateful to South West Tourism for supplying us a valuable set of data for our internal use; using this, we got encouraging results from our prototype retrieval engine. Our ideas are encompassed in several recent joint papers (see the publications list on my home page). I am pleased to thank the Leverhulme Foundation for support for this overall work and in particular for funds to employ Lindsey Ford. Ideas and work in progressThese documents reflect some of the ideas derived from our work up to now, and also cover work that is ongoing, or on a wish list. Compared with the published papers, many of them are, of course, half-baked.
Evaluation of caching:
BackgroundOriginal inspirationMy original inspiration came from a period working at Xerox EuroPARC in Cambridge during 1992-3. (It is now called XRCE.) I joined the team led by Mik Lamming, who were investigating the idea of a `memory prosthesis', and building prototype systems. The idea was to capture automatically all the events in a user's working life, and allow her to retrieve these later -- usually by context. Thus the system could become an aid to human memory, and indeed a good substitute for a filing system. As is typical of Xerox projects, this was a far-reaching and imaginative project that was possibly ahead of its time, but has spawned many follow-ups several years later. These followers stand on the shoulders both of the original Xerox researchers and of the advances in hardware over the intervening years. I started my own work in 1995, and like many of the followers, chose a more tractable problem that was a subset of the original Xerox work. Specifically I was interested in context-ware triggering, and this led to the technology called Stick-e notes. I worked with Jason Pascoe, Nick Ryan, John Bovey and David Morse at the Univ. of Kent. I have also continued the association with Xerox, and have spent another period working with them in Cambridge. We have recently -- in collaboration with authors from the USA and continental Europe -- produced a joint paper on future context-aware applications.
Stick-e notesStick-e notes are an electronic generalisation of Post-it notes: a context is attached to some information and the information is retrieved (either by proactive triggering or by normal information retrieval) when the user's current context is matched. There are several papers on stick-e notes in the publications list on my home page -- see dates from 1996 to 1999. The strength and the weakness of Stick-e notes is that they are simple. We built a number of prototypes, and these led to many insights, and, hopefully, an advance in the state-of-the art of context-aware applications. An underlying philosophy of this and subsequent work is that it is best to think of context as a many-faceted entity rather than to think of context as just location and to bolt on other contextual fields later. Some lessons learned from our work have been:
|
|||
