This repository contains two assignments developed for the Natural Language Processing course of the Master's Degree in Artificial Intelligence at University of Bologna during the A.Y. 2022/2023.
Both assignments were done in Colab Notebooks, hence it's recommended to run them in the same environment. However, there shouldn't be any problems in other IPython Notebooks, except for some limitations (e.g. mounting a Google Drive storage).
-
The objective of the first assignment is to perform Part Of Speech tagging considering different variation of a Recurrent Neural Network (LSTM). You can find more details in the relative directory.
-
The goal of the second one is to perform Abstractive QA on the CoQA dataset, using transformer-based architectures with BERT-like models for both encoder and decoder part. More information can be found in the relative directory.
In both directories you can find the notebooks that were used to explore the data, clean the data, train the models and evaluate them. Furthermore you will find the final reports that describe the work; since we had some limits for the number of pages you may look for some more information within the notebooks as well.
For both assignments we decided to use Tensorflow and Keras, moreover we used Huggingface for the QA one.
We suggest you to open the 2 notebooks on Google Colab because we inserted some forms to make the entire notebook easier to read.
These assignments have been implemented by: