I. Introduction
The design and development of Safety-Critical Embedded Systems (SCES) is a relevant activity in many application fields such as railway, automotive, aerospace, health, etc. The life cycle of a new product must follow both regulatory constraints and challenging requirements and is continuously affected by a number of conflicting objectives to be achieved such as minimization of costs, improvement of performances, short time to market. Both service levels and deadlines of commitments must be satisfied without failing the required, more and more challenging, safety constraints. Testing represents a relevant phase that both stresses the system and measures and certificates the achievement of the objectives. Beside the understanding of specification and analysis of requirements is very critical for test definition. In fact it is time consuming, it depends on the interactions of from heterogeneous stakeholders. We can identify at least the experts of the application domain and technicians, or anyone is in charge to understand high level system requirements and produce the necessary not ambiguous specification to feed the design and development process. Semantic mismatches, lack of formal models, the absence of internationally recognized open standards are limiting factors that affect the industrial performance, above all in the test definition phase. In fact they consume about the 25% of the entire testing phase [1]. Many errors occur during the test definition, when it needs to translate the high level specification to formal executable scripts, and they main reason of failures of testing, which sometime imply the revision of the project across the entire life cycle. In this paper we presents a methodology to improve performance of test definition by increasing the level of formalization of test cases and supporting the task of the project engineer with tools that use both semantic techniques and text analysis.