Distributed Cbr using XML [C. Haeys, P. Cunningam, M. Doyle ] Department of Computer Science Trinity College Dublin Conor.Hayes@cs.tcd.ie

The full version of the draft of   the paper is available in postscript format 

ABSTRACT

CBR (case-based reasoning) has considerable potential for developing intelligent assistants for the World  Wide Web. Several case-based intelligent applications already exist on the web but this applications follow a thin client model with the intelligence located at the server side. 

In this paper we explore the advantages  of making these applications more distributed . We illustrate  the kind of application where the dialogue with the case base is long lived and the network latency or server load would suggest that some of processing should be distributed to the client side.  

We present an architecture  for such a distributed CBR system and describe how a case-representation language based on XML can facilitates this distribution. The Advantages of adopting an XML case-representation language are interoperability and flexibility of reuse and these are discussed in the second half of the paper  

[Original abstract from the paper; emphases - words in bold - by maf ]

PAPER-OUT-LINE

1. Introduction 

Overview of the idea behind the on-going research:

    • Incremental CBR (I-CBR)
    • Distributed Architecture for such a system (to maximise response-time efficiency)
    • Necessity of open standards for case based network computing and interoperability with non CBR systems 

2. Web based Application 

    • Example of an internet application (I-CBR type of) with a long lived interaction with case base (Travel Agents Case-Base). 
    • Thin client systems implementation of long lived interaction application result in poor response  times.
    • Distributed CBR as solution to network latency and server load (=> causes of poor response times for the users)

3. Distributed CBR 

    • Description of an architecture for distributed CBR That allows some of the case-base processing to be distributed on the front end.
    • The CBR engine is downloaded to the client side to allow for later stages of processing to be performed on the client side
    • Contras of Java: (case-base could be passed as objects to the client, but won't work across firewalls). Proposed solution: XML based case representation [can be passed as plain text using http protocol]

4 . Case representation in XML [introduction to CBML] 

5.Advantages of CBML 

    • Application independent [data can be read by a browser and then passed to a local application for further processing ]
    • Multiple visual representations
    • Granular updating

6.Conclusions