This repo contains all the laboratory lectures for the Natural Language Understanding course held at the University of Trento.
The exercises must be delivered using the directory exam/studentID_name_surname. In this directory, you will find a folder for each lab /LAB_{1 to 9}.
From lab 1 to 8, the folders contain 3 files namely main.py, functions.py and README.md.
functions.py: you have to write all the other required functions (taken from the notebook and/or written on your own) needed to complete the exercise.main.py: you have to write the calls to the functions to output the results asked by the exercise.README.md: you may want to write a message for us related to your solution (optional).
From lab 9 to 11, the folders contain two folders one for part 1 and the other for part 2. Inside them, there are the following files and folders: main.py, functions.py , utils.py, model.py, README.md, /dataset and /bin.
utils.py: you have to put all the functions needed to preprocess and load the datasetmodel.py: the class of the model defined in Pytorch.functions.py: you have to write all the other required functions (taken from the notebook and/or written on your own) needed to complete the exercise.main.py: you have to write the calls to the functions needed to output the results asked by the exercise.README.md: you may want to write a message for us related to your solution (optional)./dataset: the files of that dataset that you used./bin: the binary files of the best models that you have trained.
Furthermore, for each of the three last exercises, you must write a small report using the LaTeX template in the zip folder NLU_final_report.zip. In particular, you have to write a mini-report of max 1 page (references, tables and images are excluded from the count). Reports longer than 1 page will not be evaluated. The purpose of this is to give you a way to report cleanly the results and give you space to describe what you have done and/or the originality that you have added to the exercise. You can find more detail about the sections and relative content in the LaTeX template.
Last but not least, the code has to be well-written and documented with comments. Furthermore, the script has to run without bugs otherwise the exercise will not be evaluated. Jupyter notebooks are not accepted.
To submit your work you have to fill out this Google form. The work must be delivered 7 days before the date of the session exam that you want to attend. You can do multiple submissions as we will check only the last one.
In the repo labs, you can find the notebooks for each lab session and in solutions you can find the same notebooks with the solutions.
The solutions of each lab will be uploaded after the corresponding lab lecture.
The notebooks that you can find here are an adaptation of the labs created by our colleague Evgeny A. Stepanov.