TY - GEN
T1 - Enhancing Sentiment Analysis with GPT—A Comparison of Large Language Models and Traditional Machine Learning Techniques
AU - Obinwanne, Tobechi
AU - Brandtner, Patrick
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.
PY - 2024
Y1 - 2024
N2 - Sentiment analysis is the process of extracting and analyzing opinions, attitudes, and emotions expressed in text data. Due to the increased amount of user-generated content on the Internet, sentiment analysis has become an important research area, which has successfully been tackled with machine learning models such as SVMs and Random Forests. Against the backdrop of rapidly growing popularity of Generative Pre-trained Transformers (GPT), the question arises, as to how good such models perform in sentiment analysis. This research paper analyzes and compares the performance of GPT-based models with traditional machine learning models for sentiment analysis. The results paint a clear picture: GPTs are a powerful concept that are applicable in sentiment analysis. In our study, they outperform traditional models such as SVM, Random Forests, or Naïve Bayes based on F1-score, precision, accuracy, and AUC-ROC score. As more advanced versions of GPT continue to be developed, it is likely that these models will become even more effective and popular sentiment analysis. Hence, the application and evaluation of GPTs represents a promising avenue for future research.
AB - Sentiment analysis is the process of extracting and analyzing opinions, attitudes, and emotions expressed in text data. Due to the increased amount of user-generated content on the Internet, sentiment analysis has become an important research area, which has successfully been tackled with machine learning models such as SVMs and Random Forests. Against the backdrop of rapidly growing popularity of Generative Pre-trained Transformers (GPT), the question arises, as to how good such models perform in sentiment analysis. This research paper analyzes and compares the performance of GPT-based models with traditional machine learning models for sentiment analysis. The results paint a clear picture: GPTs are a powerful concept that are applicable in sentiment analysis. In our study, they outperform traditional models such as SVM, Random Forests, or Naïve Bayes based on F1-score, precision, accuracy, and AUC-ROC score. As more advanced versions of GPT continue to be developed, it is likely that these models will become even more effective and popular sentiment analysis. Hence, the application and evaluation of GPTs represents a promising avenue for future research.
KW - ChatGPT
KW - Generative pre-trained transformers
KW - GPT
KW - GPT-3.5
KW - Large language models
KW - Machine learning
KW - Sentiment analysis
UR - http://www.scopus.com/inward/record.url?scp=85187702830&partnerID=8YFLogxK
U2 - 10.1007/978-981-99-7569-3_17
DO - 10.1007/978-981-99-7569-3_17
M3 - Conference contribution
AN - SCOPUS:85187702830
SN - 9789819975686
T3 - Lecture Notes in Networks and Systems
SP - 187
EP - 197
BT - Intelligent Sustainable Systems - Selected Papers of WorldS4 2023
A2 - Nagar, Atulya K.
A2 - Jat, Dharm Singh
A2 - Mishra, Durgesh
A2 - Joshi, Amit
PB - Springer
T2 - World Conference on Smart Trends in Systems, Security, and Sustainability, WS4 2023
Y2 - 21 August 2023 through 24 August 2023
ER -