In this paper we present a method for automatically testing interactive multimodal systems. The proposed approach was originally dedicated to synchronous programming which is mainly used for real-time systems. Nevertheless, the behaviour of real-time systems, consisting of cycles starting by reading an external input and ending by issuing an output, is to a certain extent similar to the one of interactive systems. So considered, this paper investigates the use of the Lutess testing environment dedicated to synchronous software for automatically testing multimodal interactive (not necessarily real-time) systems. More precisely, we focus on test data generation based on operational profiles. The main benefit of this approach is that it increases the ability to automatically test an interactive system over long input sequences, according to various use profiles.

Nowadays, test cases may correspond to elaborate programs. It is therefore sensible to try to specify test cases in order to get a more abstract view of these. This paper explores the notion of test purpose as a way to specify a set of test cases. It shows how test purposes are exploited today by several tools that automate the generation of test cases. It presents the major relations that link test purposes, test cases and reference specification. It also explores the similarities and differences between the specification of test cases, and the specification of programs. This opens perspectives for the synthesis and the verification of test cases, and for other activities like test case retrieval.