Sentence Pair Classification Fine-Tuning of Bidirectional Encoding Representation from Transformer (BERT) for Aspect-Based Sentiment Analysis (ABSA) on IMDb Movie Reviews

Aspect Based Sentiment Analysis (ABSA) provides a more detailed sentiment analysis towards aspects on domains like movie reviews, which often discussing various aspects of a movie in a review. However, previous models often struggle with handling sarcasm and other complex language nature. This study explores the fine-tuning of Bidirectional Encoder Representations from Transformers (BERT) for ABSA on IMDb movie reviews by implementing sentence pair classification that incorporates auxiliary sentences representing predefined aspects such as acting, plot, directing, visual, and audio. The dataset consists of 2,000 unique reviews expanded into 10,000 sentence pair sequences. Two models were developed: one utilizing the base BERT model directly trained on the IMDb dataset, and the other fine-tuned on SemEval 2014 dataset before application to IMDb reviews. The fine-tuned BERT model on the SemEval 2014 dataset outperformed the baseline model in low training data scenarios, achieving an F1-score of 0.88 for the aspect detection task and sentiment classification accuracy of 80%, 82%, and 89% for 4-way, 3-way, and binary classification settings respectively. he fine-tuned BERT model on the SemEval 2014 dataset also achieves more stable performance across low-resources scenarios.