Applying Text Data Analytics Techniques to Wine Reviews

Authors

Gregoire J. Thebaud, University of North Carolina WilmingtonFollow
Pierre Kraus, University of North Carolina WilmingtonFollow
Minoo Mondaresnezhad, University of North Carol?ina WilmingtonFollow

Abstract

Abstract When purchasing a bottle of wine, we usually rely on the smell and flavor descriptions from experts’ wine reviews. Just like restaurants, critics’ reviews count as a marketing strategy in the wine industry. If your wine consistently impresses the tasters and makes their top picks, there is a good chance customer will follow and buy it. A study by Friberg and Gronqvist (2012) found that the effect of a favorable review peaks in the week after publication with an increase in demand of 6 percent, and the effect remains significant for more than 20 weeks (Friberg & Grönqvist, 2012). This project is part of a student paper for a Text Data Analytics course. The goal of this study is to evaluate the description of wines provided by the tasters. The data was scraped from Wine Enthusiast (https://www.winemag.com/), a world-known American multi-channel marketer during the week of June 15th, 2017. The dataset includes fields such as the type of grapes used to make the wine, the country that the wine is from, the number of points Wine Enthusiast rated the wine on a scale of 1-100, the cost for a bottle of the wine, taster full name, description of the wine by a sommelier, etc. We used R programming to analyze the dataset. We will apply exploratory data analysis methods to find insights and interesting facts about this dataset. Specifically, we will count the number of reviews based on variety and flavors. Then, we will look at the total number of reviews by the taster. Tasters with at least 100,000 reviews can be categorized as master reviewers. We will use this knowledge to categorize reviews based on their expertise and experiences. We will also examine which variety of wine received the best score by reviewers and identify the reviewers’ favorite types of wine. We will also calculate the point distribution based on reviewers. Concerning the description column, the first step will be to pre-process and clean the text to deal with the missing values, remove the special characters, etc. We will apply descriptive text analytics techniques to understand the tasters’ reviews better. We will calculate the most frequently used words in the descriptions, as well as the most important terms based on TF-IDF. We will also apply sentiment analysis and topic modeling techniques to categorize the tasters’ reviews. Sentiment analysis enables us to extract insights from qualitative data, such as wine reviews. These insights can be used by the winemaker to respond to their clients’ needs. By detecting positive, neutral, and negative opinions within the text, we can understand how tasters feel about a particular wine and make data-driven decisions. We will identify the polarity of reviews based on the taster’s name, the varieties of wine, and the country. Polarity in sentiment analysis refers to identifying sentiment orientation (positive, neutral, and negative). Using it for our reviews could potentially yield exciting results. Finally, we will use classification methods to identify the common topics in the tasters’ wine descriptions.

Recommended Citation

Thebaud, Gregoire J.; Kraus, Pierre; and Mondaresnezhad, Minoo, "Applying Text Data Analytics Techniques to Wine Reviews" (2020). AMCIS 2020 TREOs. 53.
https://aisel.aisnet.org/treos_amcis2020/53