Multi-Class Sentiment Analysis with BERT
Fine-tuning bert-base-uncased on the SMILE annotation dataset, as an end-to-end study of transformer transfer learning.
Problem
A multi-class sentiment classification task used to work through transformer fine-tuning end to end rather than to establish a new result.
Method
Fine-tunes a pre-trained bert-base-uncased transformer on the SMILE annotation dataset in PyTorch, with a complete NLP pipeline — tokenisation, encoding into input IDs and attention masks, and DataLoader construction for batched training and evaluation — trained with the AdamW optimiser and a linear learning-rate schedule with warm-up steps.
Results
The resume records strong validation accuracy through transfer learning; a specific figure is not published here because none was recorded in a verified source.
Limitations
A learning project on a single small dataset. It is included for completeness rather than as a research contribution.