Application of an electronic tongue for evaluating basic gustatory attributes perceived in table olives: qualitative and quantitative approaches
Conference Paper
Overview
Additional Document Info
View All
Overview
abstract
The organolep:c evalua:on of table olives aiming their commercial classifica:on according to sensory trade
categories, although not being legally required, is recommended by the Interna:onal Olive Council. This
classifica:on is based on the organolep:c evalua:on of nega:ve a\ributes usually found in table olives (and
their respec:ve brine solu:ons), performed by trained sensory panels. However, the training and
implementa:on of such panels is :me-consuming, costly and has some drawbacks like the low number of
samples that can be assessed per day as well as the intrinsic degree of subjec:vity of the evalua:ons carried
out by the trained panelists. Besides the percep:on of sensory defects (type and intensity), panelists are
usually asked, among other characteris:cs, to assess the intensity of basic gustatory a\ributes (e.g., acid,
bi\er and salty), which knowledge is useful for table olives quality control. In this work, and for the first
:me, the poten:al use of a home-made electronic tongue for discrimina:ng standard aqueous solu:ons of
chemical compounds (obtained with mineral water and in the concentra:on ranges used during sensory
panels training sessions) that mimic the 3 basic tastes is evaluated: tartaric and citric acids (0.01 to 2 g/L; for
acid taste); caffeine and anhydrous quinine (0.01 to 3 g/L; for bi\er taste); and, sodium and potassium
chloride (0.5 to 25 g/L; for salty taste). The results showed that a linear discriminant model based on the
poten:ometric signals recorded by a sub-set of 5 sensors (composed by cross-sensi:vity lipidic membranes)
could correctly classify the standard solu:ons according to the basic taste mimicked with sensi:vi:es of: (i)
98% for the leave-one-out cross-valida:on; and (ii) 98% ± 3% (ranging from 91% to 100%) for the repeated
K-folds cross-valida:on (K = 4 folds with 10 repeats, guaranteeing that 25% of the original data was kept for
internal-valida:on purposes). Furthermore, the poten:ometric signal profiles recorded by the electronic
tongue during the analysis of table olives and respec:ve brine solu:ons (40 different samples) were used,
for the first :me, to quan:ta:vely es:mate the median intensity of the same gustatory a\ributes (acid,
bi\er and salty) perceived by a sensory panel (composed by 8 trained panelists) during the simultaneous
analysis of table olives and brine solu:ons. The results showed that it was possible to establish sa:sfactory
mul:ple linear regression models based on sub-sets of signals gathered during the analysis of the table
olives and/or brine solu:ons (varying from 21 to 25 depending on the basic taste), also selected by applying
the simulated annealing variable selec:on algorithm: (i) R2 ≥ 0.968 for leave- one-out cross-valida:on; and
(ii) R2 ≥ 0.97 ± 0.02 for the repeated K-folds cross-valida:on (K=4 folds with 10 repeats). These preliminary qualita:ve and quan:ta:ve results allow foreseen the prac:cal applica:on of the electronic tongue for
assessing gustatory basic tastes on table olive real samples, which could be used as a helpful tool for the
hard task required to sensory panelists.
The organolep:c evalua:on of table olives aiming their commercial classifica:on according to sensory trade categories, although not being legally required, is recommended by the Interna:onal Olive Council. This classifica:on is based on the organolep:c evalua:on of nega:ve a\ributes usually found in table olives (and their respec:ve brine solu:ons), performed by trained sensory panels. However, the training and implementa:on of such panels is :me-consuming, costly and has some drawbacks like the low number of samples that can be assessed per day as well as the intrinsic degree of subjec:vity of the evalua:ons carried out by the trained panelists. Besides the percep:on of sensory defects (type and intensity), panelists are usually asked, among other characteris:cs, to assess the intensity of basic gustatory a\ributes (e.g., acid, bi\er and salty), which knowledge is useful for table olives quality control. In this work, and for the first :me, the poten:al use of a home-made electronic tongue for discrimina:ng standard aqueous solu:ons of chemical compounds (obtained with mineral water and in the concentra:on ranges used during sensory panels training sessions) that mimic the 3 basic tastes is evaluated: tartaric and citric acids (0.01 to 2 g/L; for acid taste); caffeine and anhydrous quinine (0.01 to 3 g/L; for bi\er taste); and, sodium and potassium chloride (0.5 to 25 g/L; for salty taste). The results showed that a linear discriminant model based on the poten:ometric signals recorded by a sub-set of 5 sensors (composed by cross-sensi:vity lipidic membranes) could correctly classify the standard solu:ons according to the basic taste mimicked with sensi:vi:es of: (i) 98% for the leave-one-out cross-valida:on; and (ii) 98% ± 3% (ranging from 91% to 100%) for the repeated K-folds cross-valida:on (K = 4 folds with 10 repeats, guaranteeing that 25% of the original data was kept for internal-valida:on purposes). Furthermore, the poten:ometric signal profiles recorded by the electronic tongue during the analysis of table olives and respec:ve brine solu:ons (40 different samples) were used, for the first :me, to quan:ta:vely es:mate the median intensity of the same gustatory a\ributes (acid, bi\er and salty) perceived by a sensory panel (composed by 8 trained panelists) during the simultaneous analysis of table olives and brine solu:ons. The results showed that it was possible to establish sa:sfactory mul:ple linear regression models based on sub-sets of signals gathered during the analysis of the table olives and/or brine solu:ons (varying from 21 to 25 depending on the basic taste), also selected by applying the simulated annealing variable selec:on algorithm: (i) R2 ≥ 0.968 for leave- one-out cross-valida:on; and (ii) R2 ≥ 0.97 ± 0.02 for the repeated K-folds cross-valida:on (K=4 folds with 10 repeats). These preliminary qualita:ve and quan:ta:ve results allow foreseen the prac:cal applica:on of the electronic tongue for assessing gustatory basic tastes on table olive real samples, which could be used as a helpful tool for the hard task required to sensory panelists.
