Recipe recommendation considering the flavor of regional cuisines

Abstract

There are a variety of recipe recommendation methods based on user's preferences, nutrition balance, or user's health condition. However, there is little study on recipe recommendation considering flavor preferences of regional cuisines, which is helpful for a restaurant to plan on launching dishes from other regions, and to be well received by the local people. Therefore, we propose a method to recommend a restaurant the dishes of other regions in terms of flavor similarity among the regional cuisines in China. Firstly, we quantify ingredient preferences of a regional cuisine by TF-IDF (Term Frequency-Inverse Document Frequency) and then score the dishes of regional cuisines by ingredients preference. Secondly, the cosine theorem is used to compute the flavor similarities between regional cuisines. Thirdly, inspired by the Tidal-Trust algorithm, we compose the score of regional recipes and the flavor similarity between regional cuisines into a recommendation. Lastly, the top N dishes of other regions are recommended to a restaurant. The results of our questionnaire evaluation for the dishes recommended using the proposed method were that the mean satisfaction degree of two professional chefs is 77%, and the satisfaction degrees of 75 percent of the rest respondents are all above 70%.