6533b7dbfe1ef96bd1270243

RESEARCH PRODUCT

Comparative training procedures to learn odor descriptors: effects on profiling performance

F. SauvageotI. LesschaeveC. SulmontSylvie Issanchou

subject

Orange juice0303 health sciencesOdor perception030309 nutrition & dieteticsbusiness.industryComputer science[SDV]Life Sciences [q-bio]Pattern recognition04 agricultural and veterinary sciences040401 food scienceSensory analysisSensory Systems[SDV] Life Sciences [q-bio]03 medical and health sciences0404 agricultural biotechnologyOdorProfiling (information science)Statistical analysisFruit juiceArtificial intelligencebusinessComputingMilieux_MISCELLANEOUSFood Science

description

Three groups of ten naive assessors were recruited to perform an odor profiling of 10 orange juices using 8 odor descriptors, These panels differed on the way they learned each descriptor. Group 1R learned to match one descriptor with one external standard; group 3R learned to associate one descriptor with 3 external standards; group 0R defined by itself the 8 descriptors from a set of orange juices. Results showed that the use of one standard per descriptor seemed to be only efficient when the standard was typical of the odor perception in the orange juices. Learning one odor concept with 3 standards led to redundant use of discriminant descriptors and failed on the agreement among assessors. Finally, group 0R was a discriminant as well a homogenous panel.

https://hal.inrae.fr/hal-02694849