Sentiment Analysis in Contrast to Official Data During the COVID-19 Pandemic in Ecuador

Abstract

Ecuador was one of the first Latin American countries to have a proven case of the new coronavirus SARS-CoV-2. The social networks were the media most used by citizens to replicate news about the pandemic, and issue comments about the handling of the health crisis. This article aims to present a web tool for sentiment analysis on Twitter with three different ways to analyze the corpus and polarities: a word-dictionary-based model, a custom trained supervised machine learning model, and an open-source library to process textual data and allows obtaining a polarity metric from a tweet. Then, to define the final polarity of each tweet, an ensemble machine learning model is used for combining the predictions from the three techniques through a hard majority voting ensemble. The web system was developed with free software tools and is accompanied by visualizations and statistical graphics.