High levels of noise are a well-known source of occupant discomfort in restaurants. While much previous work has focused on assessing subjective perceptions of restaurant soundscapes through survey instruments, less attention has been given to mining online user reviews for keywords related to psychoacoustic predictors. In this study, a dataset of online user reviews for over 2,500 restaurants, bars, and coffee shops in New York City has been obtained and compared against crowd-sourced measurements of noise levels within the same establishments. Analyses of review scores and the occurrence rate of acoustically relevant keywords in review text have been conducted using machine learning techniques to suggest design thresholds for restaurants across different categorizations. Results from this work can be used by consultants and owners to design restaurant soundscapes which may be more favorably reviewed.