Analisis Sentimen Pengguna Twitter terhadap Citra Merek ChatGPT Menggunakan Rule-based, Support Vector Machine, dan Lexicon-based

Abstract

A chatbot based on artificial intelligence, ChatGPT, stirred controversy in various news media as it could potentially replace human jobs, particularly in writing. As a new technology, ChatGPT naturally faced pros and cons, leading to skepticism within the community about its brand image. Therefore, a sentiment analysis specifically targeting Indonesian-language Twitter users focusing on the brand image of ChatGPT is conducted. The research employed a combination of rule-based, Support Vector Machine, and lexicon-based methods. Rule-based classification utilized emoticons and comparatively, while lexicon-based classification used the BabelSenticNet lexicon. The dataset, acquired through Twitter crawling, initially comprised 2500 tweets, which are then cleansed to 1728 tweets. Following the classification, model evaluation is performed by comparing the performance between the proposed classification model and its constituent model. The proposed classification model outperforms, achieving an accuracy of 86,3%, precision of 87,1%, recall of 86,5%, and an f-measure of 86,6%. Predicting sentiment for 1728 tweets yielded 739 positive, 385 negative, and 604 neutral sentiments. In conclusion, ChatGPT brand image in Indonesia leaned towards positivity, notwithstanding contrasting views and objective assessments.