Next: References Up: Corporate Memories as Distributed Previous: Discussion

Conclusions

Corporate Memory forms an important enabling component in any knowledge management strategy of a company. In this paper, we propose that such memories be viewed as distributed case bases in order to facilitate exploitation of techniques from multi-agent systems and case-based reasoning to build flexible and powerful tools for knowledge access and manipulation. Negotiated Retrieval deals with piecing together partial responses from different sources to evolve a coherent response to a query. It augments CBR agents with distributed constraint optimization search capabilities to avoid harmful interactions among case pieces and assemble ``good'' candidate overall responses from partial responses of individual agents. FPL-based cooperative CBR proposes two cooperation modes among CBR agents for exploiting ``collective memories'' of peer agents. In DistCBR, an agent transmits the task to be achieved to another agent, and the host agent uses its CBR methods and its local case base to achieve the task for the originating agent. In ColCBR, an agent transmits the task to be achieved and the method to be used to achieve this task to another agent, and the host agent uses its local case base along with the received method to achieve the task for the originating agent. Both Negotiated Retrieval and FPL-based cooperative CBR are appropriate for different kinds of tasks and complement each other as tools for effective knowledge management, to be invoked for problems and tasks for which they are appropriate. As we gain further insights into the mechanisms for distributed case-based reasoning systems through our future research agenda, we hope to translate these into tools for knowledge management in the corporate memory context.


nagendra@cs.umass.edu
Mon Sep 16 17:23:45 EDT 1996