The aim of this project was to create question-answering based on deep learning that answers in style of language used in Avatar: The Legend of Aang (officially known as Avatar: The Last Airbender).
To solve the problem we used data available in Kaggle datasets, under the name Avatar: The Last Airbender (link)
To reproduce results you have to clone the repository and run indicated source files.
Quality is assured by a CI/CD pipeline consisting of the following stages:
flake8
mypy
pytest
Three of us, namely Ania, Konrad and
Witek are familiar with the cartoon, but one person, Jacek, has never watched even a single episode. Therefore, to
equalize expertise and to introduce you to the ATLA world we prepared EDA. In the jupyter notebook:
notebooks/exploratory_data_analysis.ipynb
you can find our discoveries, among others, most frequent lemmas and
similarities between characters.