by Jacob Finerty & Ofir Klein
Summary of Intent:
Given its international reach, and nearly 3 billion user-base, Facebook rebranding into Meta is anything but inconsequential. With Meta's forthcoming "Metaverse," conversations regarding "NTFs," "Web 3.0," and "Cryptocurrencies" are at the forefront of public discourse. As to what these technologies are, what they do, or even how such technologies work were all leading questions following from Meta's inception. While Meta advertised its future prospects with positivity and excitement, social media platforms like Twitter published tweets ranging from positive sentiments, fostering the excitement, negative sentiment, noting doubt and fear, and neutral sentiment, indicating, for example, an indicative mood. Focusing on the tweets concerning cryptocurrency, like bitcoin, the high-level question became: can we capture the sentiment and topics traversing tweets concerning cryptocurrency like bitcoin? Because public discourse surrounding cryptocurrency is framed either as or between positive and negative sentiment, our project intends to place sentiment at the center of the analysis by first clustering the data around sentiment scores, then move to extract the topics of the sentiment clusters.
In short, sentiment analysis provides a means to interpret and classify data with interest towards subjectivity and emotion, usually ranging from a positive to a negative score. For data saturated by sentiment, e.g., Twitter data, the question this project aims to answer is how to classify and cluster tweets by its sentiment? And, further, what topical insight can be derived from the clusters themselves? What kind of ideas are being expressed? In this work, we follow serverial research domains to construct a sentiment-cluster-topic extraction model.
Data Source: https://www.kaggle.com/datasets/226bc93769b5b28697eb8fc4a107040b3145c01640c91c16bde8424170f2820e/metadata