In ESPRIT-project nx, 2589 'Multilingual Speech Input and Output Assessment Methodology and Standardisation' (SAM), eight European languages (viz. Danish, Dutch, English, French, German, Italian, Norwegian and Swedish) form the subject of investigation. As a basic support tool for the development of assessment methods a standard approach to the (semi-automatic) labelling of speech is needed, which will serve, at the same time, the European speech community at large. In the present contribution some of the work going on in the ESPRIT-SAM project are displayed. Some of the labelling criteria used are presented, as well as some data on the agreements between labellers. To conclude, it is presented how the standard approach to semi-automatic labelling is envisaged.
