Researchers analyzed over 900,000 online restaurant reviews to understand how people structured positive and negative reviews from a narrative perspective.

Negative reviews, especially in expensive restaurants, were more likely to use features previously associated with narratives of trauma: negative emotional vocabulary, a focus on the past actions of third person actors such as waiters, and increased use of references to “we” and “us”, suggesting that negative reviews function as a means of coping with service–related trauma. Positive reviews also employed framings contextualized by expense: inexpensive restaurant reviews use the language of addiction to frame the reviewer as craving fatty or starchy foods. Positive reviews of expensive restaurants were long narratives using long words emphasizing the reviewer’s linguistic capital and also focusing on sensory pleasure. Our results demonstrate that portraying the self, whether as well–educated, as a victim, or even as addicted to chocolate, is a key function of reviews and suggests the important role of online reviews in exploring social psychological variables.

Anyone who's spent some time reading restaurant reviews on Yelp will feel a pang of recognition. Perhaps taking a restaurant to task after a poor service experience is a cathartic way of dealing with the trauma, explaining why someone might take the time to write yet another review when a restaurant on Yelp already has hundreds or thousands of reviews.

In summary, one–star reviews were overwhelmingly focused on narrating experiences of trauma rather than discussing food, both portraying the author as a victim and using first person plural to express solace in community.

The narrative style of the hyperbolic negative restaurant review, with its first person framing, has made them well-suited to serving as dramatic monologues.

On positive reviews of inexpensive restaurants:

Whether there is in fact a biochemical link between junk food cravings and drug addiction is an open question in the literature [4]. Nonetheless, our results suggest that the folk model of this belief is productive and widespread in consumer reviews. Hormes and Rozin (2010) found that participants rated the words “craving” and “addiction” in various languages as being most appropriately applied to drugs, alcohol, or food. Our study extends these results to show that the metaphor of food as an addiction or craving tends to apply to a particular subset of foods. The foods that are “craved” are foods that are in some way non–normative: they are meaty, sugary, starchy foods, generally fast food and street food, or small snack–like inexpensive ethnic foods. Craved foods aren’t vegetables, or main courses like meatloaf or fish or even side dishes like mashed potatoes. The folk model of what we crave or are addicted to encompasses foods that are somehow considered inappropriate for a meal, bad for you (unhealthily full of fats and sugars), inexpensive, comfort food that we feel guilty for having but eat anyhow.
The result that women are more likely to use this metaphor in our data is also consistent with previous results. Rozin, et al. (1991) found that females are significantly more likely to express cravings for chocolate than males. Zellner, et al. (1999), Weingarten and Elston (1990), and Osman and Sobal (2006) found that female undergraduates were more likely than males to report food cravings. Our results do not distinguish among the possible causes of the greater number of these expressions by female reviewers: women might be more likely than men to have these cravings or feelings, women might be more comfortable than men to admitting to these cravings, or women might simply be more likely than men to use this particular linguistic metaphor to describe their otherwise identical desires.

This makes intuitive sense. Most people are more likely to eat more affordable food repeatedly and thus describe them as an addictive substance. I wonder if the prevalent discussion of the health risks of of meaty, starchy foods contributes to the language of addiction when describing them; guilt and addiction are so entwined.

Positive reviews of expensive restaurants use a different narrative framing.

We first examined review features linked with educational capital. Education is strongly associated with differences in socioeconomic status, and in fact is one of the main ways that class status is defined in social scientific studies, along with work and income. Previous work on food advertising found that advertising of more expensive products employs longer, more complex words and longer sentences (Freedman and Jurafsky, 2011), presumably because complex words or sentences signal the writers’ higher educational capital, and hence project higher social status. We therefore tested whether this use of more complex language to project “linguistic capital” was similarly associated with price in reviews, predicting that reviews more expensive restaurants would be longer and use longer words.
The second feature we investigate frames food as a sensual or even sexual pleasure. This tendency is widespread in expensive wine reviews, which make extensive use of phrases like sexy, sensual, seductive, voluptuously textured, ravishing, and hedonistic (Lehrer, 2009; McCoy, 2005; Shesgreen, 2003). Television food commercials in the United States also emphasize “sensual hedonism” with words like luscious, indulgent, irresistible, and decadent (Strauss, 2005). We therefore expected reviews of expensive restaurants to use words related to sex or sensuality.

The data confirmed their hypotheses.

I wonder how much of this narrative framing results from some level of confirmation bias. If you spend so much on a restaurant meal, you're going to feel great pressure to justify your decision, and describing the meal with more more complex words and longer sentences might be one way to justify your expense as having led to a more sophisticated, sensual experience.