This code was developed by: Niklas Canova, Arjun Butani and Maria Romeiro Santos

How to run the scripts:

We have a requirements.txt file. The embeddings folder contain the transformer embeddings from exercise 3. These can be regenerated with representation_pretraining.ipynb

EXERCISE 1

• src/preprocessal.py: Note that you may need to change the folder and outcome paths to load the data correctly.
• data_exploration.ipynb: Here the file paths may need to be adjusted as well

EXERCISE 2

• 2.1.1 src/classifier_last_timestamp.py (The logs can be seen in the logs folder)
• 2.1.2 src/classifier_extracted_features.py (The logs can be seen in the logs folder)
• 2.2 (LSTM): Run Q2p2.ipynb
• 2.3a: q3_transformer.py (This is combined with a part of q3 hence the name)
• 2.3b: Transformer_preprocess.py (paths may need to be changed), Transformer_Triplets.ipynb

EXERCISE 3

• 3.1, 3.2.2, 3.3: representation_pretraining.ipynb
• 3.2.1: q3_transformer.py, last part

EXERCISE 4

• 4.1 - Run Q4p1.ipynb
• 4.2 - Run Q4p2.ipynb
• 4.3 - Run Q4p3.ipynb