Battle of Sentiment Lexicons: Wordnet, Sentiwordnet, Textblob and Vader in Web Forum Analysis

Abstract

Sentiment analysis plays a vital role in assessing public perception across different sectors, including the realm of high-rise properties. This study investigates the applicability and effectiveness of four lexicon-based sentiment analysis tools—WordNet, Textblob, SentiWordNet, and Valence Aware Dictionary and Sentiment Reasoner (VADER)—in measuring public sentiment towards high-rise properties in Malaysia. Utilizing a real-world case study, the classification performance and accuracy of these lexicon dictionaries were evaluated. The findings reveal that all four dictionaries predicted a higher proportion of positive reviews in comparison to negative and neutral reviews. WordNet recorded the highest number of positive reviews, closely followed by SentiWordNet and VADER. In contrast, VADER identified the most sentences with negative polarity, followed by WordNet and SentiWordNet. Regarding neutral polarity, VADER showed the highest count of neutral reviews, while SentiWordNet and WordNet had fewer. Given the overall performance, VADER proved to be the more effective lexicon compare to the other three. Misclassification cases by VADER exposed limitations such as restricted vocabulary, difficulties in detecting sarcasm, ambiguity, and issues with misspelled or short terms. This study offers valuable insights for researchers, practitioners, and policymakers in analyzing public sentiment toward high-rise properties, there by facilitating informed decision-making based on precise sentiment classification.