Comparative Sentiment Analysis of Cryptocurrency Apps Using BERT

Abstract

Natural Language Processing (NLP) has significantly advanced the ability to analyse and interpret textual data, playing a crucial role in understanding user sentiments. This study applies cutting-edge NLP techniques to perform sentiment analysis on reviews from India’s top cryptocurrency apps and Twitter data. Given the growing interest in cryptocurrencies, understanding user sentiment is vital for market insights and product improvement. We collected a diverse dataset from the Google Play store and Twitter, encompassing 7,197 reviews and numerous tweets. Utilizing the BERT (Bidirectional Encoder Representations from Transformers) model, known for its deep learning capabilities, we processed and analysed the data. The dataset underwent thorough pre-processing, including tokenization and the removal of irrelevant elements. Our analysis compared the BERT model’s performance with traditional classifiers such as Naive Bayes and Support Vector Machines (SVM). Findings show that BERT significantly outperformed other models, achieving superior precision, recall, accuracy and F1-scores. These results underscore the effectiveness of advanced NLP models in sentiment analysis, particularly for understanding public sentiment towards cryptocurrency apps in India. Future research will explore sentiment analysis on a broader range of platforms and fine-tuning BERT model parameters to further enhance performance and accuracy.