We introduce TIRA, a software architecture for the rapid prototyping of tailored information retrieval (IR) tools. TIRA allows to compose personal IR tools from atomic IR services, following the model driven architecture (MDA) paradigm: In a first step, an IR process is defined independently from platforms, by means of a UML activity diagram. In a second step, the activity diagram is transformed to a platform specific model, which is executed in a distributed environment.

Major driving force behind our research is the question of personalization: We see a large gap between information retrieval theory and algorithms on the one hand and their implementation and deployment to satisfy a personal information need on the other. This gap can be closed with adequate software engineering means, and TIRA shall contribute in this respect.