This repository stores the code developed during the NLP course at FEUP. It is divided into project 1 and project 2, which target the same objective and address a classification task over Portugues argumentative text. The first part uses only classical machine learning techniques. The second part uses hugging face transformers to address this problem.
Text spans that are taken to have an argumentative role (either as premises or conclusions of arguments). There are 5 possible labels in this multi-class classification problem. Value, Value(+), Value(-), Fact, or Policy.
This project was an appendix to the DARGMINTS project whose teaching staff is an active part of it. It uses news from the Portuguese newspaper Público.
The first approach performed better but was tougher to achieve, transformers provided a straightforward approach to this problem but were harder to manage.
Code developed by students at FEUP. Use it in your own responsibility
Inês Quarteu up201806279@up.pt
Rúben Almeida up201704618@up.pt