VADER: A Parsimonious Rule-Based Model for Sentiment Analysis of Social Media Text

The paper presents VADER, a parsimonious rule-based model for general sentiment analysis aimed at the challenges posed by social media content. The authors first construct and empirically validate a gold-standard list of lexical features, along with their associated sentiment intensity measures, specifically attuned to microblog-like contexts. They then combine these lexical features with five general rules that embody grammatical and syntactical conventions for expressing and emphasizing sentiment intensity.

VADER is compared against eleven state-of-practice benchmarks, including LIWC, ANEW, the General Inquirer, SentiWordNet, and machine-learning techniques based on Naive Bayes, Maximum Entropy, and SVM. Using the model to assess the sentiment of tweets, VADER outperforms individual human raters, achieving an F1 classification accuracy of 0.96 versus 0.84 for humans, and it generalizes more favorably across contexts than any of the tested benchmarks. This showed that a simple, interpretable rule-based approach could rival more complex machine-learning methods.