Abstract: In the context of high quality food, such as meals in a gastronomic restaurant, classical approaches using liking levels hardly differentiate recipes as products are all rated excellent. More, recipes are often complex, with many stimuli contributing the overall judgment: the visual dressing of the plate, the flavour and texture of each component, e.g. meat, vegetable 1, vegetable 2, sauce, starch… The objective of this work is to define a way to measure consumer opinion of complex dishes allowing to compare recipes and draw conclusions for possible improvements. The selected method is based on spontaneous language as it is hypothesised that emotions would be more relevant than liking scores to reach a better discriminability. Consumers are asked to give 3 words to describe their experience of the food. Then the R3M algorithm takes into account the language structure: the meaning but also the world nature (grammatical, logic, valence...) and the verbalisation context. Consumers are also asked to rate their liking of the food and give positive and negative reasons explaining the given scores. Regular clients of the restaurant participated in the study (n=120). Four dishes of two menus were evaluated : Menu Créativité and Menu Bien-être. In a general perspective, the two methods gave consistent results and allow to show the menu Créativité was slighty more preferred than the other one. In a detailed analysis, the spontaneous language approach allowed to identify associated emotions that were not verbalised in the liking measures. More specifically, it also allowed to identify some sensory characteristics to be adjusted for better satisfaction. The spontaneous verbalization helped to go beyond rational diagnosis, bringing emotional dimensions crucial for this high-end culinary universe, such as surprise, aesthetics and originality of the dishes.
