Analyzing Cultural Assimilation through the Lens of Yelp Restaurant Reviews

Abstract

Given the steady stream of immigrants from around the world, cultural assimilation in North America has long been a topic of interest. However, existing research focuses only on assimilation to North American culture, overlooking the mutual influence, with a very limited use of data-driven approaches. In this paper, we investigate assimilation among various cultures in North America through the lens of discussions surrounding food. We first present Cross-Cuisine Cross-Region LDA (c3rLDA), a novel probabilistic graphical model to jointly uncover latent topics shared across cuisines, as well as their regional variants for each cuisine. Then, we employ the model on 3.7 million Yelp restaurant reviews to find that cuisines assimilate to one another in varying degrees depending on the cuisines involved, the topic, and the region: A cuisine tends to be more influenced by other cuisines if it is regularly fused with others (e.g. Japanese), for certain topics (e.g. breakfast and dessert), and in specific regions (e.g. stronger Mexican influence in the Southwestern US and French influence in the East Canada). Lastly, we demonstrate that the topics generated by our model, on which the qualitative analysis is based, are more coherent than or comparable to those generated by existing neural and non-neural topic models. This work represents the first step toward large-scale data-driven analysis of cultural assimilation in North America, which is made possible by the abundant data available in social media.