Accuracy over time frankish et al 1992 on the other

This preview shows page 6 - 8 out of 30 pages.

accuracy over time [Frankish et al. 1992].On the other hand, recognition accuracy is not the only determinant ofuser satisfaction. Both the complexity of error recovery dialogues [Zajicekand Hewitt 1990] and the value-added benefit for any given effort [Frank-ish et al. 1995] affect user satisfaction. For example, Frankish et al. [1995]found that users were less frustrated by recognition errors when the taskwas to enter a command in a form than when they were writing journalentries. They suggest that the pay-back for entering a single word in thecase of a command is much larger when compared with the effort ofentering the word in a paragraph of a journal entry.Error handling is not a new problem. In fact, it is endemic to the designof computer systems that attempt to mimic human abilities. Research inthe area of error handling for recognition technologies must assume thaterrors will occur, and then answer questions about the best ways to dealwith them. Several research areas for error handling of recognition-basedinterfaces have emerged:Error reduction:This involves research into improving recognition tech-nology in order to eliminate or reduce errors. It has been the focus ofextensive research, and could easily be the subject of a whole paper on itsown. Evidence suggests that its holy grail, the elimination of errors, isprobably not achievable.34G. D. Abowd and E. D. MynattACM Transactions on Computer-Human Interaction, Vol. 7, No. 1, March 2000.
Error discovery:Before either the system or the user can take any actionrelated to a given error, one of them has to know that the error hasoccurred. The system may be told of an error through explicit user input,and can help the user to find errors through effective output of uncertaininterpretations of recognized input. Three techniques are used to auto-mate such error discovery—thresholding of confidence measures, histori-cal statistics [Marx and Schmandt 1994], and explicit rule specification[Baber and Hone 1993].Reusable infrastructure for error correction:Toolkits provide reusablecomponentsandaremostusefulwhenaclassofcommon,similarproblems exists. Interfaces for error handling would benefit tremen-dously from a toolkit that presents a library of error-handling techniquesof recognition-based input. Such a toolkit would have to handle theinherent ambiguities that arise when multiple interpretations are gener-ated for some raw input. A prototype toolkit has been proposed byMankoff et al. [2000] to support reusable recovery techniques, but manychallenges remain.3. CONTEXT-AWARE COMPUTINGTwo compelling early demonstrations of ubicomp were the Olivetti Re-search Lab’s Active Badge [Want et al. 1992] and the Xerox PARCTab[Want et al. 1995], both location-aware appliances. These devices leveragea simple piece of context, user location, and provide valuable services(automatic call forwarding for a phone system, automatically updated mapsof user locations in an office). Whereas the connection between computa-

Upload your study docs or become a

Course Hero member to access this document

Upload your study docs or become a

Course Hero member to access this document

End of preview. Want to read all 30 pages?

Upload your study docs or become a

Course Hero member to access this document

Term
Spring
Professor
GregoryD.Abowd

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture

  • Left Quote Icon

    Student Picture