Employing supervised algorithms, such as Naïve Bayes, sentiment analyses were performed for 1,182 English reviews for 12 Michelin Bib Gourmand restaurants in Phuket on TripAdvisor. The algorithm generated 68 terms with a frequency of over 50. We manually divided the words into four groups based on conventionally accepted attributes for measuring dining perception theory and the conclusion of a word list from the literature reviews presented in Table 1. Words that were irrelevant to the four dining dimensions were grouped as others (Table 2). The words in the “others” group were allocated a vector that contained the emotion found in the lexical resources for emotions in the English language (Giatsoglou et al., 2017). These words were loaded in the utilized emotion dimensions regarding sentiment polarity, as well as the spectrum of eight emotions by Plutchik (1994), namely anger, anticipation, contempt, fear, pleasure, grief, surprise, and trust. Although the words in this group contain significant values for defining the sentiment polarity, this study mainly focused on terms related to restaurant dimensions as the main priority. Thus, only the words in the four attributes for measuring dining perception/experience were employed to determine significant differences in the parameters of the terms used in reviews among restaurant dimensions using an ANOVA analysis. With the question, “Is there a difference in the means of the salience and valence parameters of four different restaurant attributes?” The study’s null hypothesis (H₀) posits no significant difference in customers’ perception (salience) and satisfaction (valence) regarding the four attributes under investigation, measured by the significance level at 0.05 of ANOVA. The results are presented in Table 3.