Diagnosis based on sensory data: Application to wheat grading quality

Abstract

Sensory evaluation is an important aspect of food quality and control. However, even when carried out by a group of experts, it is generally difficult to link the results of a sensory evaluation to physico-chemical or technological measurements. This study is based on the premise that formalising the interpretation of sensory observations in terms of the physical state of the product can help to link together sensory and physical properties. The main proposal of this paper is a methodological framework adapted from a diagnostic approach to capture the relationships between sensory evaluations of a type of product, here wheat dough, and its physical states called quality profiles. A probabilistic analysis is proposed to identify the quality profiles and their signatures, i.e. the corresponding sensory observations that result from grouping the probabilities of the observations. This work is supported by the analysis of a large historical sensory evaluation dataset from the routine application of the French baking standard to estimate the baking value of common wheat (Triticum aestivum L.) flour. Application of the method to this dataset revealed two defective quality profiles for wheat dough, Slackening (due to weakness of the gluten network) and Resistant (excessive strength of the gluten network), along with their signatures in terms of sensory observations of the dough. Promising relationships were found between the quality profiles attributed to the wheat samples and usual technological criteria of the wheat flour quality: gluten index, (Ie) elasticity index and (W) dough strength. This methodological framework applied to food opens up interesting perspectives for the use of sensory data for crop and food quality assessment using computational approaches.