Carnot Intelligent Agents and Digital Libraries

Darrell W. Woelk

MCC, 3500 West Balcones Center Drive, Austin, Texas 78759,

1. Introduction

The Carnot Project at MCC is addressing the problem of logically unifying physically-distributed, enterprise-wide, heterogeneous information. Carnot will provide a user with the means to navigate information efficiently and transparently, to update that information consistently, and to write applications easily for large, heterogeneous, distributed information systems. A prototype has been implemented which provides services for (a) enterprise modeling and model integration to create an enterprise-wide view, (b) semantic expansion of queries on the view to queries on individual resources, and (c) inter-resource consistency management. Carnot also includes technology for 3D visualization of large information spaces, knowledge discovery in databases, and software application design recovery.

2. Carnot Intelligent Agents

The Carnot system is implemented as a set of distributed, communicating intelligent agents. For the integration of heterogeneous databases, these agents are automatically programmed with the knowledge necessary to map among different data models, query languages, and database schemas. This knowledge is created as the result of an integration phase in which a new database is described to Carnot in terms of its relationship to the concepts in an enterprise model. When a query is received, an agent is then dispatched to the remote site where a database resides. The agent retrieves the requested data from the remote database using the correct query language, transforms the data where necessary, and returns the properly formatted data [3]. Carnot agents have been developed for various DBMSs (Ingres, Oracle, Sybase) and various standard APIs and protocols (SQL Access Group Call Level Interface, ISODE Remote Database Access protocol).

We are also developing intelligent agents for accessing textual databases. These agents will support protocols such as Z39.50 and WAIS for keyword-based text retrieval queries. Additionally, since Carnot can maintain an enterprise model in a knowledge base, it is possible to transform keyword-based text queries into queries against structured databases, such as relational databases. We have also developed some capability for translating natural language queries into queries against structured databases.

The Carnot intelligent agents are implemented in the Rosette language. Rosette is a prototype-based, object-oriented command language based on the Actor model that incorporates sophisticated support for the dynamic definition of foreign language storage structures and type-checked access to procedures generated by foreign language processors and linked dynamically into the runtime environment [2]. These features make it possible to rapidly integrate new facilities into Rosette and to configure software based on Rosette at runtime in the field. Rosette actors executing on different systems communicate through the Rosette Tree Space. The Tree Space is an enhanced implementation of the Linda Tuple Space communication paradigm [1].

3. Carnot and Digital Libraries We believe that the capabilities that have been developed in the Carnot project at MCC are applicable to accessing information in digital libraries. For example, just as an enterprise model is valuable in the integration of enterprise databases, a Carnot model of an information domain can help to provide intelligent indexing to concepts in a digital library. Furthermore, the Carnot intelligent agents can provide an excellent flexible, extensible environment for distributed searching of libraries. A Carnot agent can also be dispatched to a remote site where it can monitor the information being added to a library. When information is added that matches a certain criteria, the agent can alert a user to this event through the Carnot Tree Space.


[1] Carriero, Nicholas and Gelernter, David. 1989. Linda in Context. CACM Volume 3, Number 4.

[2] Tomlinson, C., Attie, P., Cannata, P., Sheth, A., Singh, M., and Woelk, D. 1993. Workflow Support in Carnot, In Data Engineering Bulletin, Vol. 16, No. 2, June, 1993, pp. 33-36.

[3] Woelk, D., Cannata, P., Huhns, M., Shen, W., and Tomlinson, C. 1993. Using Carnot for Enterprise Information Integration. In Proceedings of Second International Conference on Parallel and Distributed Information Systems. January, 1993. pp. 133-136.

Last Modified: