Best Restaurant in Boston

[by Naveen]

What's the best restaurant in Boston?
I saw a recent article in Boston Magazine that claimed to have the answers. They compiled results from a variety of sources - the Boston Globe, Herald, and Phoenix, as well as Chowhound, Yelp, and the Phantom Gourmet - into "one single, convenient, yet statistically meaningful index." I was intrigued by this endeavor, but was more interested in the reverse question: what are the main disagreements among the reviewers? I was especially curious about the differences among the critics. For instance, Devra First from the Globe is a "self-styled gastro-populist," while the Phantom Gourmet loves "Fried things, bad puns, 'ooey-gooey' desserts, heaping piles of...anything."

I figured that the best way to pursue this question would be to turn it into my final project for my computer science class about data visualization. At first I was overwhelmed by the mismatch of letter grades, starts, and numbers used by the different reviewers in different years, but Prof. Elaine Allen at Babson College was generous enough to give me the normalized scores from each review source. I used a computer language called Processing (a Java descendent) in order to create an interactive visualization. At first I followed my instincts as a physict and made a two-dimensional scatter job with different-sizes of color-coded dots, but when the post-docs in my lab took a while to figure out what was going on, I knew I needed a different approach. After eight iterations, I ended up with a set of four bar charts, as shown below:

The actual visualization is interactive and should get posted to the course website soon, but for now you'll just have to imagine all types of color-changing and highlighting awesomeness. The composite score goes from lowest to highest along the horizontal axis, with the top-ranked restaurant on the right. The length of the bar shows how much higher or lower a particular critic ranked that restaurant. Red bars don't indicate a negative review - they just mean that this reviewer ranked the venue lower than average. By applying the various filters on the right-hand-side, it's possible to find interesting patterns, such as:
  • Best-value restaurant: highest ranked for the lowest cost [41. Garden at the Cellar]. The $$ filter is great for grad students who want the occasional gastronomically-inspiring experience, but can't afford a trip to a $$$$ restaurant.
  • Worst-value restaurant: lowest ranked for the highest cost [81. Mooo]. This is just the result from some quasi-arbitrary numbers and I this is by no means a dis-recommendation.
  • Personal favorite: lowest composite score for the 25th best according to a particular critic [Boston Magazine: 63. Myers+Chang; Globe: 68. blu; Herald: 86. Bricco; Phoenix: 77. da Vinci]. If I had more time I would be more systematic about this to look at the content of each reviewer's comments for the reviews with the greatest discrepacny, perhaps using a word cloud.
  • Popular favorite: lowest composite score for highest-Chowhound rating [104. Grotto]. There's also Yelp, Urban Spoon, and many other sources that I didn't have time to investigate. Dealing with all of Yelp.com's idiosyncrosies in URL name by hand was difficult enough.
I wouldn't take any of these numbers too seriously, since there are numerous things that people seek when eating out. A few examples that come to mind:
  • The perfectly executed version of a favorite dish (e.g. Thomas Keller's quest for perfection)
  • An innovative meal that challenges expectations: a few courses may be unpleasant, but at least it will be unlike anything eaten before (e.g. Grant Achatz's cuisine)
  • Cuiosity about a hot new restaurant and the ability to be a trend-setter (an expensive hobby for most people)
I have a vision for a ranking system that would rate restaurants based on the goals of the diner, their personal reference points, and ratings from like-minded critics. To me, Malcolm Galdwell's analogy of spaghetti sauce comes to mind: initially there was just one type of sauce that every manufacturer made. Then, one company hired a consultant who realized that some people prefer thin sauce, while others like chunky. Now we the supermarket shelves are packed with dozens of customized vareities. I think that on-line rankings are just beginning to realize their full potential.

What are your thoughts? I'd love to hear about epxeriences at any of these top restaurants in Boston or ideas for the visualization.

No comments:

Post a Comment