Using Large Language Models for Aspect Based Sentiment Analysis
Author
Term
4. semester
Publication year
2024
Submitted on
2024-06-02
Abstract
This thesis investigates the use of Large Language Models (LLMs) for Aspect-Based Sentiment Analysis (ABSA). We start with traditional machine learning techniques and the move on to advanced deep learning models. The models were assessed based on their effectiveness in category classification, sentiment polarity classification, and joint classification tasks. Fine-tuning experiments with OpenAI's GPT-3.5 Turbo model showed significant improvements in performance, highlighting the benefits of model adaptation. Despite these advances, limitations such as data availability, computational resource requirements, and model hallucinations were identified. The study recommends future work on expanding datasets, optimizing model architectures, and exploring advanced fine-tuning techniques to enhance the robustness and accuracy of LLMs for ABSA in real-world applications.
Documents