These are four models for sentiment classification on tweets, developed for the Artificial Intelligence 2 (Deep Learning for Natural Language Processing) course at UoA's Department of Informatics and Telecommunications, taught by Manolis Koubarakis, during the Spring Semester of 2024-2025. Each model was developed and optimized separately using a different method. The same training set, validation set and test set was used for all models. Experiments were performed both with the model's architecutre and hyper-parameters, as well as the data pre-processing. For the first model, a data analysis was also performed on the training and validation set before and after the text pre-processing, to study the effect it had on the data.
Each model is accompanied by a report detailing the experimentation process, analyzing the results and comparing each model with the previous ones. The BERT and DistilBERT models were developed together and share a report.