AI-POWERED TEXT ANALYSIS TOOL FOR SENTIMENT ANALYSIS

Abstract: In today’s digital era, text data plays a ubiquitous role across various domains. This bachelor thesis focuses on the field of sentiment analysis, specifically addressing the task of classifying text into positive, negative, or neutral sentiments with the help of an AI tool. The key research questions addressed are: (1) How can an accurate sentiment classification system be developed to categorize customer reviews as positive, negative, or neutral? (2) How can the performance of the sentiment analysis tool be optimized and evaluated, considering the factors that influence its accuracy? (3) How does Chat-GPT evaluate text-based feedback from customers with our results as input, i.a. can"Artificial General Intelligence" be adapted to solve a specific problem in the domain of this work? To accomplish this, the study harnesses the power of RoBERTa, an implemented transformer model renowned for its prowess in natural language processing tasks. The model will mainly focus on review comments from Amazon and on the product, "Samsung Galaxy A53". A small comparative analysis will also be carried out with Chat-GPT and the RoBERTa model’s sentiment positions. The results demonstrate the effectiveness of the RoBERTa model in sentiment classification, showcasing its ability to categorize sentiments for different review comments. The evaluation process identified key factors that influence the tool’s performance and provided insights into areas for further improvement. In conclusion, this thesis contributes to the field of sentiment analysis by providing a comprehensive overview of the development, optimization, and evaluation of an AI-powered text analysis tool for the sentiment classification of customer reviews. The result affects the importance of understanding customer sentiment and providing practical implications for businesses to improve their decision-making processes and enhance customer satisfaction.