Sei Abe, Takahiro Matsui, Shige Koseki, Kento Koyama
{"title":"Hierarchical ranking sensory evaluation test of fresh produce quality: A case study of injured cherry tomato images","authors":"Sei Abe, Takahiro Matsui, Shige Koseki, Kento Koyama","doi":"10.1016/j.foodqual.2024.105167","DOIUrl":null,"url":null,"abstract":"<div><p>Sensory evaluation is used to identify how humans perceive and respond to stimuli. However, current judgment criteria are limited to the number of ordinal categories. These ordinal categories do not account for variations within them, making it hard to grasp the panelist's detailed preferences fully. This study aimed to develop a hierarchical ranking test to reveal human response. The quality of 100 cherry tomato images was ranked using a joint paired comparison and ranking test. One hundred images of cherry tomatoes with visible damage were investigated for sensory evaluation. Twelve untrained panelists evaluated the quality of the cherry tomato images, ranking them from 1st to 100th. As a result, the Kendall rank correlation coefficient, which evaluates the agreement of ranking among panelists, was 0.84, suggesting that the evaluation of the quality of cherry tomatoes generally showed the same trend among 12 panelists. For all 12 panelists, the proportion of the wounded area on the tomato had a more significant impact on the quality ranking than the wound area size. In addition, the quality ranking of cherry tomatoes was predicted by fine-tuning the convolutional neural network model VGG-19 with a moderately high coefficient of determination of 0.82. Our new hierarchical ranking test would facilitate detailed human evaluations, upon which ranking predictions would be made.</p></div>","PeriodicalId":322,"journal":{"name":"Food Quality and Preference","volume":null,"pages":null},"PeriodicalIF":4.9000,"publicationDate":"2024-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Food Quality and Preference","FirstCategoryId":"97","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0950329324000697","RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"FOOD SCIENCE & TECHNOLOGY","Score":null,"Total":0}
引用次数: 0
Abstract
Sensory evaluation is used to identify how humans perceive and respond to stimuli. However, current judgment criteria are limited to the number of ordinal categories. These ordinal categories do not account for variations within them, making it hard to grasp the panelist's detailed preferences fully. This study aimed to develop a hierarchical ranking test to reveal human response. The quality of 100 cherry tomato images was ranked using a joint paired comparison and ranking test. One hundred images of cherry tomatoes with visible damage were investigated for sensory evaluation. Twelve untrained panelists evaluated the quality of the cherry tomato images, ranking them from 1st to 100th. As a result, the Kendall rank correlation coefficient, which evaluates the agreement of ranking among panelists, was 0.84, suggesting that the evaluation of the quality of cherry tomatoes generally showed the same trend among 12 panelists. For all 12 panelists, the proportion of the wounded area on the tomato had a more significant impact on the quality ranking than the wound area size. In addition, the quality ranking of cherry tomatoes was predicted by fine-tuning the convolutional neural network model VGG-19 with a moderately high coefficient of determination of 0.82. Our new hierarchical ranking test would facilitate detailed human evaluations, upon which ranking predictions would be made.
期刊介绍:
Food Quality and Preference is a journal devoted to sensory, consumer and behavioural research in food and non-food products. It publishes original research, critical reviews, and short communications in sensory and consumer science, and sensometrics. In addition, the journal publishes special invited issues on important timely topics and from relevant conferences. These are aimed at bridging the gap between research and application, bringing together authors and readers in consumer and market research, sensory science, sensometrics and sensory evaluation, nutrition and food choice, as well as food research, product development and sensory quality assurance. Submissions to Food Quality and Preference are limited to papers that include some form of human measurement; papers that are limited to physical/chemical measures or the routine application of sensory, consumer or econometric analysis will not be considered unless they specifically make a novel scientific contribution in line with the journal''s coverage as outlined below.