Sentiment Analysis of User Perceptions Toward the Gemini Application using the Naïve Bayes Algorithm

Abstract

Sentiment analysis, a pivotal technique in Natural Language Processing (NLP), facilitates the classification of textual data into positive or negative sentiment categories. This study applies a Naïve Bayes classification model to analyse 1,000 user reviews of the Gemini application, an AI-powered assistant developed by Google AI, collected from the Google Play Store. The research adopts the Knowledge Discovery in Database (KDD) methodology, encompassing data selection, preprocessing, transformation, data mining, and evaluation stages. Reviews were pre-processed using text normalization techniques and transformed into numerical vectors using Term Frequency–Inverse Document Frequency (TF-IDF). The sentiment classifier achieved a high accuracy of 94.57%, with a precision of 95.7% and an F1-score of 96.7%, indicating effective classification performance. The results revealed a predominance of positive sentiment, with frequently occurring terms such as “assist,” “good,” and “application.” Conversely, negative sentiments were often associated with issues regarding language support and system limitations. These findings provide actionable insights into user satisfaction and areas for improvement in the Gemini application. While the Naïve Bayes algorithm proved efficient and interpretable, limitations include difficulty handling nuanced linguistic features such as sarcasm or ambiguity. Future research may enhance sentiment classification performance by incorporating larger datasets and advanced models such as BERT or deep neural networks. Overall, this study demonstrates the practical utility of machine learning-based sentiment analysis in guiding user-centric application development.