Sentiments comparison on Twitter about LGBT
Aldinata, Axell Mondrian Soesanto, Vincent Christian Chandra, Derwin Suhartono
Abstract
Twitter has become one of the most popular mediums for people to voice their opinions freely on any topic from around the world. Most companies use sentiment analysis from Twitter to see how well their product is. These opinions are categorized into three types which are positive, negative, and neutral sentiments. This paper focuses on Sentiment Analysis of LGBT, which has become a trending and controversial topic for society to talk about. We took these Tweets from fifty states in the United States of America and pre-process them before classifying their sentiment. We test five algorithms to classify the sentiments which are TextBlob PatternAnalyzer, Naive Bayes, Linear Support Vector Machine, Logistic Regression, and XGBoost using both non pre-processed data and pre-processed data. We found that the Logistic Regression without text-preprocessing gives us the best result with a 70.87% F1-score. After applying our sentiment classifier to the US tweets that talked about LGBT, we found that most tweets have a neutral sentiment.