5 Conclusion

 
Several approaches have been proposed for querying multiple heterogeneous information sources using mediator-based architecture, such as TSIMMIS [2], Information Manifold [4], DISCO [1], and the context interchange project [10]. The key aspect distinguishing DIOM from other systems is its strive for logical independence (e.g., query independence and source independence). We implement such notion by (1) identifying the key aspects of the user query profile that are useful for capturing the semantic context (and scope) of what the user wants in a specific query, (2) by supporting the content and capability description of an information source independently of the user queries and of other information sources, and (3) by offering mechanisms for dynamically interconnecting the information sources to a query through the query routing process. Another contribution of the paper is the semantic relevance reasoning strategies we use to discover and prune the information sources that do not contribute to a query. A novel aspect of our relevance reasoning methods is the step-wise utilization of the user query profile and the source capability descriptions to ensure that each step considerably reduces the number of candidate sources considered in the next step.

The first prototype implementation of DIOM distributed query processing system (DQS) [9] was developed and tested on Solaris platform using SunJDK version 1.1. The byte-code has been tested on the following platforms: Windows NT v.3.5.1 using Netscape Navigator v.2.01, and Sun OS v.4.1.4 using Netscape Navigator v.3.0.

Our future research will examine the role of source capability and query profile metadata in automating the query result packaging and assembly process. We believe that in an open environment, it is more beneficial to resolve the hard heterogeneity problems by providing semantic attachment mechanisms to annotate the results returned from heterogeneous sources, rather than using the enforcement of a pre-defined integration approach.