Application of Traditional and Deep Learning Algorithms in Sentiment Analysis of Global Warming Tweets

Dragana Nikolova, Georgina Mirceva, Eftim Zdravevski

Abstract: The Earth’s surface is continuously warming, changing our planet’s average balance of nature. While we live and experience the impacts of global warming, people debate whether global warming is a threat to our planet or a hoax. This paper uses relevant global warming tweets to analyze sentiment and show how people’s opinions change over time concerning global warming. This analysis can contribute to understanding public perception, identify misinformation, and support climate advocacy. This paper proposes a data processing pipeline encompassing traditional and deep learning based methods, including VADER, TextBlob, Doc2Vec, Word2Vec, LSTMs, to name a few. The extensive testing shows that the combination of document embeddings and neural networks yields the best results of up to 97% AUC ROC and 93% accuracy. The findings enable the comprehension of human attitudes and actions related to this worldwide issue in production environments.

External IDs:dblp:conf/goodtechs/NikolovaMZ23