In this work, after analyzing and preparing the given corpus, we developed a series of neural architectures according to the assignment specification. After that, we selected the two best models based on the validation F1 score and evaluated them on the test set. Finally, after tuning some hyperparameters and adding regularization to the baseline model, we evaluated it on the test set and we obtained an accuracy of 92.89% and an F1 score of 85.8%.
| Name | Username | |
|---|---|---|
| Lorenzo Pratesi | lorenzo.pratesi2@studio.unibo.it | Prahtz |
| Martina Rossini | martina.rossini3@studio.unibo.it | mwritescode |
| Riccardo Foschi | riccardo.foschi4@studio.unibo.it | snifus |
| Vairo Di Pasquale | vairo.dipasquale@studio.unibo.it | vairodp |