How to sound more professional in your wine reviews

Rachel Woods
The Wine Nerd
Published in
2 min readMar 28, 2019

Do you ever read some of the professional wine reviews and just laugh at how ridiculous they sound? Or read them and are thoroughly impressed by how they can match the taste of wine to something so eclectic as “wet gravel”?

I was curious about what really differs between professional wine reviews and those written by amateurs (you and me). To answer this question, I collected wine reviews professional written from Wine Spectator’s Top 100 wines’ of 2018. I then collected up to three amateur reviews for the same wines from Vivino, to get a final sample of 100 professional reviews, and 185 amateur reviews.

First, how much do professional versus amateurs write in their wine reviews?

After removing stop words (aka filler words like “the”, “it”, “and”), we compared the length of amateur wine ratings versus professionals and discovered that amateur reviews are both longer and shorter than professionals. This suggests that professionals tend to have their cadence down — they don’t write too little, nor too much about a wine.

Next, we looked at what words were used most frequently by professionals and amateurs.

Surprisingly, there aren’t huge differences in the most common words used between the two groups. Besides the common use of superlatives amongst amateurs (good, nice), there’s a lot of similarities between the groups. It’s common to use words like finish, aromas, tannins, and acidity. Professionals may talk a little more prominently about the palate and aromas than amateurs. However, there may be more difference in the long tail of words used.

In conclusion:

  • The ways that amateurs describe wines versus professionals is not as different as expected.
  • The largest different looks to come from the variability in length of reviews.
  • Vocabulary between the two groups is not that different. However, we don’t account for the differences in the long-tail of words used between the two groups. This could be the topic of a future analysis.

Code and data shared on data.world

--

--