The project focuses on enhancing automated product summary generation and sentiment analysis using advanced AI and machine learning, particularly transformers. Building on previous research, the aim is to refine the system with pre-training, fine-tuning, and an encoder-decoder structure. From summer 2023 to May 2024, the project incorporated the OpenAI API to test new technology and build on previous group work. The project's timeline included key phases like research, implementation, integration, and evaluation, emphasizing the practical application of sentiment analysis and key point extraction to lay the groundwork for comprehensive product summaries.
Using the summarized data from previous work, sentiment was analyzed and verified with BERT to ensure output quality. This method involved rigorous data preprocessing and annotation to maintain high accuracy in sentiment labeling. By leveraging BERT's capabilities, the project aimed to produce precise and insightful sentiment analysis, facilitating better decision-making processes. The approach underscores efficiency, utilizing the OpenAI API and BERT for a broader, more nuanced analysis of textual data, ultimately enhancing the product review summarization framework and offering deeper insights into consumer sentiment
For more detailed information see the report.