This reposiory contains two assignments and the project work done for the Natural Language Processing course of Alma Mater Studiorum. It is dividen in three sections: Assignment 1 (POS), Assignment 2 (QA), Project Work in Human Value Detection. It is the result of the teamwork of Alessandro Lombardini, Giacomo Melacini, Matteo Rossi Reich and Lorenzo Tribuiani.
Part-of-speech (POS) tagging is a popular NLP process that consists in marking up every word in a corpus as corresponding to a particular part of speech. This assignment aimed at ex- perimenting with the problem from the prepro- cessing to the model architectures and error analysis. This work enabled the use of different RNN models to effectively perform labeling. The experimented techniques lead to good re- sults and allowed a deep understanding of a simple NLP workflow.
Generative Question Answering (QA) is a particular task of NLP which aims to creates models and architectures capable of producing answers to given questions in the form of free text. The aim of this assignment is to create a QA system based on the CoQa dataset by fine-tuning the distilroberta-base and bertiny models from Huggingface.
Based on the Human Value Detection challenge 2023, this project discusses the implementation of NLP models capable to classify Human val- ues on which a specific text relies on. This task uses a set of 20 value categories compiled from the social science literature and described in the paper Identifying the Human Values behind Arguments (Kiesel et al., 2022).